TO MEASURE IS TO KNOW! Henrik Gadd

TO MEASURE IS TO KNOW! Henrik Gadd

TO MEASURE IS TO KNOW! 2012 Henrik Gadd Printed in Sweden by E-husets Tryckeri, Lund 2012 ISRN LUTMDN/TMHP-12/7077-SE ISSN 0282-1990 ii

ABSTRACT District heating is a mature and widely spread technology and it is an effective way to increase energy use efficiency and decrease primary energy use which is a requisite for obtaining a sustainable future. In order for district heating systems to be competitive, high concentration of heat demand is, in certain respect, the most important factor. Customers located next to the district heating system must perceive district heating as the most attractive heat supply system. This implies that the cost of heat has to be relatively low compared to other heating alternatives. This is accomplished by a combination of low cost for heat supply and high overall system efficiency. Traditionally, district heating systems have consisted of heat generation plants, district heating networks and substations. But the heat is not useful until the buildings have a comfortable indoor temperature and hot water in the taps. This is the reason why the total system should also embrace heat supply systems in customer buildings. Thanks to automatic meter reading systems new opportunities have arisen. As the title of this work says: To measure is to know, and with more than 30 000 values/year and district heating substation, it should be possible to know a lot more about the customers than how much to charge for one month of delivered heat. This new knowledge can be a first step to future smart heat grids. By using measure values from heat meters, daily heat load variations in district heating networks and heat load patterns in substations have been analysed. It is a theoretical study inasmuch as confirmation in the field is not performed. But the heat meter values used in the analyses are from actual district heating networks and actual heat supply to substations from the day to day work at a district heating company. The major findings in this work are: iii

Daily variations in district heating systems are low. Only about 5 % of the heat is above the daily average heat load in annual basis. The district heating network has both less daily and seasonal heat load variations than the aggregated heat loads of the customers. Daily variations are to a large extent social heat loads and are spread over time. Geographical diversity is also decreasing daily heat load variations in district heating networks. Seasonal heat load variations in district heating networks are less compared to substations because heat loss from networks is a significant heat demand, about 10% of the supplied heat. It is possible to identify heat load patterns in customer categories, but there are no standard heat load patterns and the variation among different customers varies greatly. This work has come at the right time because there is a large amount of metering values easily accessible, waiting to be explored. iv

ACKNOWLEDGEMENTS First of all, I would like to thank my supervisor Sven Werner at Halmstad University with whom it is a privilege to work with. He is always ready to take time to share his almost infinite knowledge of energy systems in general and of district heating in particular to answer my never-ending wonderings and questions. I would also like to thank my colleagues at Halmstad University for good company and interesting discussions around the coffee table. In particular, I would like to mention my always enthusiastic and helpful PhD-student- brother Urban Persson, who has always been a great supporter! I would like to thank the staff in the Department of Energy Sciences at LTH. My co-supervisor Svend Frederiksen for being there as a safe back -up, research secretary Elna Andersson for fast and correct administrative support and the Head of the department Bengt Sundén for his nonbureaucratic and easy manners in the cooperation between Halmstad University and LTH. I would also like to thank Fjärrsyn, the Swedish district heating research programme, for the financial support and the reference group for supportive meetings. I also thank my employer Öresundskraft for financial support and that I have got the opportunity to pursue PhD studies. In particular, I would like to thank my manager, Lars-Inge Persson, for his incessant support and optimism. At last but certainly not the least, I would like to thank my family, my wife Johanna and my three lovely boys Oscar, Mauritz and Otto, for their support even if they have not fully understood the greatness of district heating. (yet!) v

List of publications This dissertation is based on the following papers, referred to in the text by their Roman numerals. The papers are appended at the end of the dissertation. Paper I Paper II Daily heat load variation in Swedish district heating systems. Gadd H, Werner S Submitted for publication (2012). Heat load patterns in district heating substations. Gadd H, Werner S Submitted for publication (2012). Other related publications by the author Daily heat load variation in Swedish district heating systems. Gadd H, Werner S Earlier version of Paper I presented at the 12 th International Symposium on District Heating and Cooling, 2010, Tallinn. My contributions to the publications Paper I I performed the calculations and wrote the paper with guidance from Sven Werner. Paper II I developed the method together with Sven Werner. I performed the calculations and wrote the paper with guidance from Sven Werner. vi

CONTENT INTRODUCTION... 1 BACKGROUND... 1 PURPOSE AND SCOPE... 5 LIMITATIONS... 6 OUTLINE OF DISSERTATION... 7 DISTRICT HEATING TECHNOLOGY... 8 HEAT SUPPLY PLANTS... 8 DISTRICT HEATING NETWORKS... 9 DISTRICT HEATING SUBSTATIONS... 9 HEAT METERS AND AUTOMATIC METER READING... 10 BUILDING HEAT DEMANDS AS HEAT LOADS... 14 HEAT LOAD VARIATIONS... 17 HEAT LOAD VARIATIONS... 17 HEAT STORAGES... 20 METHOD... 23 RESULTS... 24 REGARDING PAPER I... 24 REGARDING PAPER II... 24 vii

CONCLUSIONS AND FURTHER WORK... 26 REFERENCES... 28 viii

INTRODUCTION Background District heating is a centralised system supplying customers with heat by distributing hot water in a pipe network, a district heating network. The first district heating system in Sweden was put into operation in 1948 [1]. Today, 2012, there are more than 200 district heating companies with over 400 district heating systems in Sweden and almost every city has a district heating system. In 2009 the district heating systems delivered about 180 GJ or 50 TWh of heat which represent 60% of the heat demand in buildings for space heating and domestic hot water. In premises and multi-dwelling buildings the market share was 80% and 92% respectively [2]. District heating market penetration is substantial mainly in northern and Eastern Europe, but in most countries in the cooler temperate climate zones of the world district heating is used. The countries with the largest market penetration are, according to [3], the Nordic and Baltic states. Hence, district heating is a technology that is mature and widely spread. Both from an environmental, locally and globally, and from an economic point of view there are several benefits from district heating as a heat supply system. The centralised heat generation has moved the combustion of fuels from local boilers in the cities with air pollution as a result to larger units outside cities with an advanced air pollution control resulting in a better local environment. Local heat sources like industrial excess heat or geothermal heat would be impossible to utilise without district heating networks to distribute the heat to the heat demands in the buildings. Municipal waste and waste wood can, in combined heat and power plants (CHP), generate useful heat and electricity with an efficiency of over 90% [4]. What used to be waste problems have become, thanks to district heating systems, useful assets while at the same time more valuable fuels are set free for other usage. This is described in for example: [5], [6], [7] and [8]. Hence, district heating is an effective way to increase efficiency in 1

energy supply and decrease primary energy use which is a requisite for obtaining a sustainable future. Customers must perceive district heating as the most attractive heat supply alternative. Large investment costs are related to heat generation plants and district heating network construction. High concentration of heat demand, high linear heat density, is in certain respect the most important factor for district heating systems in order to be competitive [9]. High linear density increases heat sales and decreases heat loss per sold energy unit and metre district heating pipe. In order to obtain high linear density and to attain competitive district heating systems, most of the buildings along the district heating network must choose district heating for their heat supply. The heat sources for district heating systems must be available at low cost since the sum of fuel cost, production and distribution must be less or, at least, equal to competing heating alternatives. This is illustrated in Figure 1. The heat production cost includes fuel and capital cost for heat generation plants or boilers. Cost Alternative heating Heat production cost District heating Distribution cost Figure 1. Illustration of the market conditions for district heating by a comparison of the cost profile for district heating and alternative heating. 2

A district system traditionally consists of three major parts. One or more heat supply plants, a district heating network for distribution of heat and, at the location of the customers, district heating substations where heat is transferred to the building heating systems and domestic hot water systems. But also heat supply systems in buildings should be considered as part of the district heating systems since customers do not buy district heating, they buy comfort. Comfort in this case implies comfortable indoor temperature and hot water in the taps. The interface between district heating companies and heat users in Sweden can be organised in several different ways. The customer is normally related to the party who signs the contract and is invoiced for the heat. The simplest relationship is for detached houses where the house owner, heat user and customer is the same person. In multi-dwelling buildings, heat is normally included in the rent, i.e. the landlord pays for the heat and is responsible for the heat supply in the building and the tenants are heat users. In larger complexes of multi-dwelling buildings with several buildings in a single property, heat can be supplied either by one district heating substation in each building or by a district heating substation in one building with a secondary supply system to the other buildings. Premises can be rented out with or without heat. These are just a few examples of possible interfaces to illustrate the complexity of a rather heterogeneous undefined group that in daily speech is called customers. If the whole system, including substations and secondary heating systems, does not work well there are two risks for the district heating company: The first risk is that investments in the district heating system is based on an incorrect heat demand/heat load pattern with a less competitive system as a result. Secondly, if the customers or heat users do not experience a proper heat delivery, and since simplicity is an important selling point for district heating, the customer may be unsatisfied with the heat delivery quality, i.e. there are at least two reasons for district heating companies to help customers use the proper amount of heat and to have a proper heat load pattern. If this is not fulfilled, in the worst case, the unsatisfied customer will change their heating system and less linear density, and an oversized heat generation capacity and an oversized district heating network will be the result. This can end up in a vicious circle of decreasing competitiveness. Actions taken at customers initiative will still probably result in decreased heat sales, but on the other hand the risk of that heat being sold will decrease. Decreased heat demands from existing 3

customers will also create room for possible expansion of new customers or new applications without new investments in increased capacity in heat generation and district heating network expansion [10]. A lot of research and investigation has been performed in order to optimise district heating and distribution. A few examples of systems overall analyses are to be found in [11], where an analysis of how district heating expansion can improve overall fuel efficiency is performed. A method to evaluate heat supply by using economic, environmental and technological factors of merit is described in [12]. Primary energy savings by using heat storages in district heating systems is analysed in [13] and a model for optimising lower costs using a data-model by applying the cost structure of district heating and electricity is presented in [14]. Regarding research in network operation, [15] presents a theoretical model of the dynamics in a district heating network. A method to use nonlinear model predictive controller to perform economical optimisation is presented in [16] and in [17] a method of cost reduction by optimising the frequency of temperature adjustments in supply temperature to the district heating network is presented. In most cases, it is taken for granted or, at least, it is not an issue that is taken into consideration that the customers heat loads/heat load patterns might not be correct. There are some works carried out regarding fault detection in district substations like [18] and [19] where heat meters are used for fault detection. Work has also been completed to analyse customers heat demand like in [20] where 50 buildings have been analysed with the focus on heat power and domestic hot water preparation, in [21] where a single multi-dwelling building is analysed focusing on thermodynamic performance, and in [22] where 8591 substations have been analysed regarding the chilling of district heating water. A decreased heat loss in customers service pipes using control strategy for a by-pass in service pipes is analysed in [23]. A most interesting conclusion is found in [24] where meter readings from electrical meters are used for analyses. Employees with knowledge in database handling and employees with knowledge of energy are not the same individuals, and this fact is delaying the development of using meter readings for analysis purposes. One hot topic on the agenda at present is smart power grids. There is no official definition of smart grids. Traditionally, electrical power distribution has been a one-way supply of electric energy from centralized power stations. Customers receive electricity at the end of the power lines and a bill in the mailbox to pay for the used electricity. In smart grids there 4

is information which is shared between the producers and the users momentarily. Electricity will, because of small scale power generation, not only be delivered but also go in the other ( wrong according to energy traditionalists!) direction, i.e. from the former customers towards the power grid. The idea to involve all the concerned parties to actively take part in a network is not unique for electrical power grids. In [25] several examples of techniques for smart cities are presented. One example is where a Google maps traffic app, instead of building a costly supervision system, uses a large number of volunteers whose mobile devices report their geographical position. Thereby information of where traffic is stopped, slow or floating can be revealed. Another example is from Copenhagen where measurements of temperature, humidity, noise and air pollution data are performed by meters mounted on bicycles communicating via the telephone network. To be competitive henceforth as is discussed in [26], district heating systems, just as other energy supply systems, need more information input. One step is to systematically detect errors in customer heating systems or control settings. The eventual errors and control settings in the customers heat demand are transferred to the heat generation plants through the district heating network, so before too much effort is put into optimisation of heat generation and distribution, errors in customers heat demands should be identified and solved. Previously maintenance has been performed periodically with visits to customers within a certain period of time. Smarter would be to have condition controlled maintenance where the condition of customers substations and heating systems is the trigger of the maintenance demand and not the calendar. This has until recently been too expensive, but with automatic meter reading systems now installed in most district heating systems in Sweden today this is achievable. All substations and heat generation plants are monitored for energy, flow, supply and return temperature and the measuring values are easily accessible at low cost. This is a first step towards smart heat grids. Purpose and scope The objective of this dissertation is increased district heating systems overall efficiency. It is most probable that the competition of heat sources will increase in the future. District heating companies must meet this challenge and retain competitiveness, and increase the efficiency of the total heat supply system. This includes the customers heat supply systems 5

since the heat demand in the heat generation plants is the aggregated heat demands of all heat customers. Two aspects in district heating systems have been investigated: Characterisation and quantification of heat load variations in district heating supplies. Heat load patterns in district heating substations. Regarding daily heat load variations in district heating systems, lots have been published on how to handle and predict heat load variations on a daily basis, but nothing has been published on how to neither characterise nor quantify daily heat load variations in district heating systems. Paper I presents a method to characterise and quantify heat load variations and an evaluation of 20 Swedish district heating systems was performed. Regarding heat load characteristics in district heating substations new opportunities have arisen since nowadays heat deliveries commonly are measured and collected hourly in an automatic meter reading system. Large quantities of measuring values are thereby easily accessible. The objective in one research project is to detect errors at the customers end by analysing the measuring data in an automatic way. Paper II presents a method using two descriptive parameters, Annual relative daily variation and Seasonal heat load variation, in order to evaluate heat customers by using measuring values from heat meters. An analysis of 141 buildings that had their heat supplied by district heating was performed. Limitations This work is based on heat measuring values from heat meters. In Paper I, hourly measured supplied heat to district heating networks is used as input data for analysis. In Paper II, hourly measured delivered heat to substations is used for analysis. It is theoretical studies inasmuch as no field verifications have been performed, but it is actual customers/heat plants with actual heat demands/supplies that have been analysed. In this study only heat energy measurements have been used and no economic evaluations have been performed. 6

Outline of dissertation The first part of the dissertation forms the background of why the studies presented in the appended papers are relevant and what have been performed in the area by other researchers. In the next chapter a description of district heating technology can be found. This part includes heat meters, an automatic meter reading system and a description of buildings heat demands as heat loads. This is followed by brief descriptions of method, results and a concluding discussion with suggested further research. Papers are appended at the end of the dissertation. 7

DISTRICT HEATING TECHNOLOGY The classic way to divide district heating systems is in three parts: Heat supply plants, district heating networks and substations. Heat from one or more heat sources e.g. heat plants like CHP-plants (Combined Heat and Power) or boilers, industrial excess heat, geothermal or solar heat is distributed in the district heating networks to each customer and transferred to the secondary systems at the customers in the district heating substations. In the substations heat is transferred to the buildings heating systems and heat for domestic hot water. District heating substations is often where heat meters are situated. Today many district heating companies have installed automatic meter reading systems due to changes in the law regarding electricity power delivery. Since most energy companies in Sweden both supply electricity and district heating, joint meter reading systems are often installed with a possibility to read all electrical and heat meters hourly. Detailed further information about district heating technology in addition to what is described below can be found in [27]. Heat supply plants Heat in district heating systems is recycled or generated in one or more heat supply plants. It can be CHP, heat only boilers or heat pumps. A variety of fuels can be used, for example, wood chips, peat, coal, oil or natural gas. But there are also possibilities to use recycled heat sources like: municipal waste, industrial excess heat or geothermal heat. What heat sources that are to be used is determined by local conditions and depends on what is locally available and most suitable. Utilization of local assets, often difficult fuels or excess heat, is one of the key parameters in the district heating systems business idea. In order to handle the fact that the heat load is changing during the year a mix of different heat generation sources are normally used. Base load heat supply plants with often high investment costs and low operational costs like waste incineration plants and peak load plants with high operational costs like oil-fired boilers. 8

District heating networks In most district heating networks in the world, water is used as heat carrier but also steam is used, mainly in the USA. The maximal design supply temperature varies between countries. In Eastern Europe the maximal design supply temperature has been about 150 C, in Germany 130-140 C, in Sweden and Finland 120 C, and in Denmark 90-120 C. The differences are partly because of tradition, partly for rational reasons. The difference between Denmark compared to Finland and Sweden can be explained by the fact that in Denmark, direct connection is the most common, but in Sweden and Finland indirect connections are most common [27]. District heating networks are what generate the flexibility in the district heating systems. A large number of customers with a heat demand connected to one distribution system creates a possibility to have a centralised heat supply where different heat sources, as described in the section above, can be used. The only demand on the heat or fuel is to be able to increase the temperature of the water in the return pipes. If each building has its own heat generation like a boiler or heat pump, it is in most cases associated with high costs to change heat generation systems and the customers are thereby committed financially when a heat generation system is installed. There must be a large difference in costs between the present heat and the alternative to make it economically competitive to change heating alternative. But in a centralized system the relative cost becomes less about changing sources of supplied heat and as is described above, there are possibilities to have different heating alternatives depending on the heat supply demand. District heating substations In district heating substations, heat is transferred from the primary system, the district heating network, to the secondary heating systems and to domestic hot water preparation. There are several connection alternatives that exist for substations. Two major groups are direct and indirect connection. Direct connection means that the water in the district heating network is directly connected to the building heating systems. In Sweden, indirect connection is the most common. The buildings secondary systems for heating and domestic hot water preparation are separated from the district heating water by heat exchangers. 9

The substations contain heat exchangers for radiator, ventilation and domestic hot water preparation, but also control systems including control valves. This is also where the heat meters are normally situated. On the primary side, the differential pressure between supply and return pipes drives the water through the substations. On the secondary side, circulation pumps are used to circulate the water in the heating systems. Domestic hot water is on the secondary side driven through the substations by the incoming cold water pressure and is not dependent on pumps within the buildings. Heat meters and Automatic meter reading Heat meters have traditionally been installed for billing purpose. The meters were read manually once or a few times a year. According to changes in the Swedish law relating to electric power delivery [28] as from 1 st of July 2009, electric power customers shall be billed for actual used electricity and not as it used to be, for preliminary estimations based on previous usage. The change in the law was a direct response of an EU-directive [29] regarding energy efficiency for end-users, stating that electricity customers should be billed for the actual use and not base the bill on previous usage. A very common situation in Sweden is that both the district heating system and the electric power grids are owned by the local energy companies which thereby are responsible for electric power metering. This is why automatic meter reading systems have been installed that are able to read both electricity meters and heat meters. From these systems hourly measures of heat energy, flow, supply and return temperature are available from all district heating customers. This opens up new opportunities. As the title of this work says: To measure is to know, and with 30 000 values/year and district heating customer, it should be possible to know a lot more about the customers than what they should pay for the last month s used heat at the end of the month. Heat meters A heat meter consists of three major parts: a flow sensor, a pair of temperature sensors and a calculator. The flow sensor and the temperature sensors are normally mounted on the primary side with one temperature sensor in the supply pipe and flow sensor and the second temperature sensor in the return pipe. Figure 2 shows a schematic heat meter. 10

T 1 Supply pipe C Return pipe T 2 V Figure 2. Schematic heat meter. T 1 =Supply temperature, T 2 =Return temperature, V=Flow sensor and C=Calculator. The energy is simplified calculated by: Q = V ρ c ΔT (1) where: Q = Energy V = Water volume ρ = Density of water volume depending on the temperature c = Specific heat capacity of water depending on the temperature ΔT = T 1 -T 2 = Difference between supply and return temperature The most common flow sensors on the market today, for district heating and cooling application, are ultrasonic sensors. Earlier, other techniques have been used and are still in use since the operation time for heat meters is up to 10 years. Examples of other previous common flow sensor technologies are impeller and magnetic-inductive flow sensors. The volume flow is measured either by flow rate dependency or constant sampling frequency. In flow rate dependency flow sensors a pulse is sent to the calculator for a fixed volume for which used heat is calculated 11

according to (1). In time dependent volume measurement, pulses are sent at a constant sampling time and the heat is calculated by a certain frequency. The sampling time is usually between 4 and 30 seconds [30]. The most commonly used temperature sensors are Pt 100, but Pt 500 and Pt 1000 are also in use. Temperature sensors are always delivered in calibrated pairs since the differential temperature, and not the individual temperatures, is used for energy calculations and thereby is the most important [31]. The calculator determines used heat energy by readings from flow meter and temperature sensors connected to the calculator and inbuilt values for density and heat capacity factors depending on the water temperature. The calculator is also equipped with a built-in meter value storage capacity. Heat meters normally deliver four parameters: volume, energy, supply temperature and return temperature. Automatic meter reading systems Commercial systems for automatic meter reading for collection of electric and district heating metering values have been in use for over 20 years. Mainly large customers have been read previously because of large costs for the systems in the past. It was economically indefensible to install automatic meter reading systems for every customer. Demands of monthly read usage of electricity combined with a decreased cost for computer systems opened up opportunities for installing automatic meter reading systems not only for electricity customers but also for district heating customers. The system consists of meters with a communication unit, and concentrator units where customers in an area are collected, as well as a central system where the metering values are collected in a metering database, see Figure 3. From the database, the values are accessible for further use like billing and analysis. The central system is where the administration, including the upgrading of all parts of the whole system, takes place. The communication between the different parts can be wired by IP, PLC (Power Line Communication), or specific communication wires or wireless by GPRS, GSM telephone network or radio. 12

Central system including metering database Concentrator Concentrator Concentrator Heat meter Heatmeter Heatmeter Heatmeter Heatmeter Heat meter Heatmeter Heatmeter Heatmeter Heatmeter Heat meter Heatmeter Heatmeter Heatmeter Heatmeter Figure 3. Schematic configuration of an automatic meter reading system. Automatic meter reading systems open up new opportunities because metering data are easily accessible in large amounts. Previous studies had to decide to either select a few objects to study and fill them with measure equipment or to study only single parameters with a larger amount of objects. This is illustrated in Figure 4 below where the lines symbolise the same cost for the study. Automatic meter reading has made it possible to move in the direction of the arrows towards the upper right corner for district heating substations. 13

10 9 8 7 6 Number of objects 5 4 3 2 1 0 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2 Number of parameters Past Present Future Figure 4. Number of studied objects versus number of parameters. Automatic meter reading in combination with decreasing cost for data collection and storage open up opportunities to evaluate a large number of parameters with a large number of objects at low cost. The lines represent equal cost. A large amount of information is now gathered at low cost on district heating customers and since the infrastructure of metering values is built up, it is easy to increase either the resolution of the existing metering or to include other variables. At present what is left to do is to dig in this gold mine and make the information useful. The limitations at the present time is not having access to metering data, but rather not having the knowledge of what to do with it. Building heat demands as heat loads Heat demands in buildings have two dimensions: where heat demands occur, and how heat demands occur. The first dimension, where heat demands occur, consists of two major heat demands: space heating and domestic hot water. The second dimension, how heat demands occur, also consists of two major groups, physical and social heat demands. Physical heat demands depend on physical conditions in terms of outdoor temperature, climate, and physical conditions of the building envelopes. Social heat demand depends on social behaviour of the people in the buildings. In Table 1 an illustration of different heat loads related to the two heat load dimensions can be found. 14

Table 1 Different heat demands related to the two heat demand dimensions: Social and Physical heat demands and Space or Domestic hot water heat demands. Social heat demands Physical heat demands Space heating Domestic hot water -Indoor temperature demand -Ventilation(time) -Shower -Dishes -Outdoor temperature -Climate -Ventilation(outdoor temperature) -Condition of building envelope -Cold water temperature Basically the amount of heat energy that needs to be added to the buildings for heating purposes is proportional to the temperature difference between the set point of the indoor temperature and the outdoor temperature. Space heat is supplied by ventilation air and radiators. The heat supply to the radiators compensates for the heat conduction through the building envelope while the ventilation air heating increases the temperature of the ventilation air to a desired supply temperature. Both these heat loads are physical heat loads since the heat demand is directly dependent on the difference between the outdoor temperature and the indoor temperature set point. But, depending on the activity in the building, the ventilation is not in operation all the time in all buildings. An office or a school does not normally require ventilation during nights and weekends since no or few people are present in the buildings at nights and weekends, in other words, ventilation is partly a social heat demand. It is both a physical heat demand since it is dependent on the outdoor temperature and a social heat demand since the ventilation system is only, or more correctly should only be, in operation when the buildings are occupied. Radiator heating compensates for the heat flow through the building envelope caused by the difference of the indoor and outdoor temperatures. When the ventilation is turned off, the radiator system prevents the building 15

from cooling off. If, for example, part of a building is not in use and the ventilation is shut off in this part, it will be kept warm by the radiator system. Heat demand for domestic hot water is only present with domestic hot water usage, i.e. domestic hot water is a highly intermittent heat demand. It is mainly a social heat demand, but since the incoming water temperatures are not constant over the year, hot water preparation has a physical heat load component. There are two techniques of domestic hot water preparation: direct with a heat exchanger or via a hot water accumulator. With direct domestic hot water preparation, large heat power peaks will occur in the hot water tap. In the heat exchanger the domestic hot water is momentarily heated by the district heating network water, whereas a hot water accumulator is a container for domestic hot water. At hot water tapping, hot water is used from the hot water accumulator and replaced with cold water. The water is then heated but not as fast as in the case with direct heated domestic hot water. In this case, the heat demand will be stretched over time and therefore not cause as large heat demand peaks as the case is with direct domestic hot water preparation. 16

HEAT LOAD VARIATIONS No work has been found on universal characterisation or quantification of daily heat load variations in district heating systems or district heating substations. In this work, heat load variations are used in two time scales. Heat load variations on a daily basis, Annual relative daily variation, and heat load variations on a seasonal basis, Annual relative seasonal variation. That heat load variations cause problems in district heating systems is obvious not at least since plenty of research on how to handle heat load variations is carried out. The orientation of the research to handle heat load variations is often either prediction of heat loads as in [32], [33], [34] and [35], to use heat storage as in [13] and [36] or to introduce Demand side management as in [37] and [38]. Heat storage in combination with demand side management is analysed in [39]. Heat load variations The origin for heat load variations in district heating systems is heat load variation at the customers substations connected to the district heating network. The origins of daily variations are mainly social heat demands such as hot domestic water and time clock operation control of ventilation. In spring and autumn, heat load variations with physical heat load origin occur as well. Differences in outdoor temperature between night and day in combination with solar incident radiation in day time increase daily heat load variations. But as can be observed in Figure 5 and Figure 6 daily heat load variations in district heating networks are far less compared to connected customers. The district heating network smoothen daily heat load variations at a large extent. The reason is an effect of geographical diversity in combination with the fact that hot water tapping, and starts and stops of ventilation within or with one another are not coordinated, i.e., there is a spread in time and hydraulic distance of peak loads in different substations. Seasonal heat load variations are likewise less for district heating systems than for the aggregated seasonal variation for the district heating 17

customers, but not as marked as with the daily variations. This fact is not as obvious as the daily variations. The probable main reason is the heat losses from the district heating network, which is part of the measures of the heat supply to the district heating systems. Heat losses from the network are more or less constant during the year. The seasonal heat load variation for this heat load is close to zero, which thereby decreases the total seasonal heat load variation in the district heating networks. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Annual relative seasonal variation Figure 5. The correlation between Annual relative daily variations and Annual relative seasonal variations for 20 Swedish district heating systems. 18

Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% Annual relative seasonal variation Continuous Night setback Time clock operation 5 Time clock operation 7 DH-systems Helsingborg Ängelholm Figure 6. The correlation between Annual relative daily variation and Annual relative seasonal variation for 141 analysed buildings with different heat load patterns. The orange oval represents the district heating systems from Figure 5 and the two large dots are the district heating systems where the analysed buildings are situated. 19

Heat storages Heat storages in district heating networks can be used to eliminate heat load variations. Everything with a mass can be used as heat storage. The storage capacity is: Q = m c ΔT (2) where: Q = Stored heat energy m = Heat storage mass c = Specific heat capacity of heat storage mass ΔT = Temperature difference in heat storage In Paper I the heat storage sizes necessary to eliminate heat load variations are stated. With a size of 17% of the daily average heat supply, 99% of the daily heat load variations can be eliminated. To eliminate seasonal heat load variations, a storage size of about 25% of annual supplied heat is required. But heat storage can be used for other reasons than eliminating daily heat load variations for example, increasing utilisation of industrial excess heat and electricity generation in CHP plants. Below, three alternative heat storage solutions are briefly described: Water tank storages, Heat storage in district heating networks, and To use heavy buildings connected to the district heating systems as heat storage. Water tank heat storages. Large water tanks in heat plants are perhaps what first comes to mind when talking about heat storages in district heating systems. The size needed in order to eliminate the heat load variations is as mentioned above 2.5 m³/tj supplied heat. But the heat storages built are, in most cases, several times larger than that. In Figure 7, the heat storage size for 41 district heating systems is plotted. It is obvious that the elimination of daily variations is only part of the reason to have heat storages in the district heating networks. To eliminate seasonal heat load variations a water storage of 1500 m³/tj supplied heat would be required. 20

Figure 7. Heat storage sizes for 41 district heating systems in Sweden. The red line is the storage size that is needed in order to eliminate 99% of the daily heat load variations. District heating networks The district heating network contains a large mass in the form of its water. In normal operation the supply temperature is reversed proportional to the outdoor temperature as can be studied in Figure 8. If the supply temperature is increased compared to what is necessary for the momentary heat supply, illustrated by the dotted line in Figure 8, then heat is stored in the district heating network. The amount of stored heat is proportional to the temperature difference between the actual supply temperature and the requisite temperature for the outdoor temperature. 21

Figure 8. Supply and return temperature as a function of outdoor temperature. If the supply temperature is increased to a level higher than the curve, i.e. to the dotted line, a certain amount of heat is stored in the district heating network proportional to the increased temperature difference. Buildings connected to the district heating networks Heavy buildings with a large mass have high time constants. This is the reason why heavy buildings can be used as heat storages without the people inside feeling less thermal comfort. By increasing the temperature in the building just a few tenths of a degree during a low-load period in the district heating network and then, during a high-load period, decreasing the heat supply to the building, it is possible to spread the supply of heat to the building over the day and thereby decreasing the daily heat load peaks in the district heating system. An evaluation of the potential of this method is described in [40]. 22

METHOD All analyses in this work are based on hourly measured heat energy from heat meters. They are theoretical analyses, but the heat meter values are from live heat meters used in the day to day work. In Paper I, the outgoing heat supply to district heating networks from the heat generation plants are used for analyses, and in Paper II heat meter values from heat meters in customer substations are used. In Paper I, Daily heat load variations are defined with a descriptive parameter called Annual relative daily variation, and 20 district heating networks are evaluated. In Paper II, two descriptive parameters are used for analysis. Annual relative daily variation, the same parameter used in Paper I and Annual relative seasonal variation defined in Paper II. Analyses are carried out on 141 buildings. The two parameters, and thereby the method, are general and not dependent on system size, and would be possible to use in other applications than in district heating. It would be possible to be used in a lot of applications where variations occurs e.g. electrical power grids, railway systems, roads and telephone networks. 23

RESULTS This section summarises the results from the appended papers. All heat load variations, daily heat load variations and seasonal heat load variations are related to the annual supplied heat energy to the district heating network in Paper I and the delivered heat to substations in Paper II. Regarding Paper I Daily heat load variations in Swedish district heating systems are small. In this study, daily heat load variations are estimated to between 3 and 6% with an average of 4.5%. Seasonal variations are at the analysed district heating systems 17 to 28% with an average of 24%, i.e., seasonal heat load variations are in the magnitude of 5 times larger than the daily variations in the analysed district heating systems. The size of heat storage to eliminate daily heat load variations is determined to be about 17% of the daily average heat supply which corresponds to 2.5 m³/tj of annually supplied heat if the storage medium is water with a temperature difference of 40 K. Loading and unloading capacity for heat storage should be about half of the annual average heat load. Regarding Paper II In customer substations, seasonal heat load variations are 20 to 40% while daily heat load variations are 5 to 25%. Seasonal heat load variation is the most dependent on customer category, type of activity in the building where industrial, commercial and public administration buildings have the highest seasonal heat load variations, Health and Social Services buildings are at an intermediate level, while multi-dwelling buildings have the lowest heat load variation. The most important cause for high daily variations is time clock operation control of ventilation. This is, or should be, implemented in 24

buildings where activities take place only parts of the day, for example in schools and offices. 25

CONCLUSIONS AND FURTHER WORK With focus on district heating systems optimisation, daily heat load variations in district heating networks and heat load variations on daily and seasonal time scales in district heating substations have been analysed and used as tools in the optimisation process. Daily heat load variations in district heating systems are small compared to seasonal variations and the spread between different district heating systems is, regarding daily and seasonal heat load variations, much less than the spread in single substations. While heat customers form a heterogeneous group, district heating systems constitute a rather homogeneous group. The district heating network has a large stabilizing effect on the heat load variations from the customers. Large daily variations with the individual customers are aggregated in the district heating network. But because of geographical diversity, and the fact that a large part of the daily heat load variations originate from social heat demands, and thereby are spread over time, the aggregated heat load variations in district heating systems are much lower compared to the daily variations at the customers substations, i.e. large heat load variations at substations do not result in large heat load variations in the district heating networks. No standard heat load pattern for the heat supply of buildings exists. This can be observed in Figure 6 where the 141 analysed buildings are presented. Within the customer categories some heat load pattern can be identified even if the variations are in a wide range. This makes it possible to identify outliers with a disadvantageous heat demand. The most important cause for different heat load patterns is time clock operation of ventilation since it causes large daily heat load variations. Analyses of heat load variations and heat load patterns in substations are rare and one reason is the previous lack of access to measuring data. In order to access measuring data in the past, measuring equipment had to be installed in selected objects. The cost of measuring was the deciding factor against performing large-scale analyses. But since nowadays automatic 26

meter reading systems are installed by many district heating companies, a large amount of metering data are now easily accessible at low cost. One of the difficulties in doing this work has been that no methods, and almost no references in research material can be found. Heat meter data used to analyse a large number of district heating customers substations on a large scale is a white spot on the map of knowledge. The overall goal for further work should be to strive towards a zeroerror vision for entire district heating systems including: heat generation plants, district heating networks, substations and heat supply systems in customer buildings. In this study, heat energy values have been used to determine two descriptive parameters. In order to automatically identify defect heat load patterns other descriptive parameters should be identified. Today also flow, supply and return temperatures of primary district heating delivery are available in the automatic meter reading systems. Technically many more parameters could be collected, e.g. differential pressure on the primary side in the substations, supply and return temperatures in radiator, ventilation and hot water circuits. One example of a descriptive parameter that would be of interest would be to identify high relative heat users, i.e. buildings with an overuse of heat. Perhaps the most important conclusion in this work confirms a wellknown fact: To measure is to know! 27

REFERENCES [1] Werner S, The heat load in district heating systems. Doctoral thesis, Chalmers University of Technology, Göteborg 1984. [2] EI R2010:04, Uppvärmning I Sverige 2010 (Heating in Sweden 2010). The Energy Markets Inspectorate, 2010. [3] Euroheat & Power, DHC_2009_Statistics_Table.pdf, District Heating and Cooling-2009 Statistics. Available at: http://www.euroheat.org/statistics-69.aspx (2012-06- 04) [4] IEA, Advancing Near-Term Low Carbon Technologies. The international CHP/DHC Collaborative, CHP/DHC Country Scorecard: Finland. Available at: http://www.iea.org/g8/chp/profiles/finland.pdf (2012-06-04) [5] Persson U, Werner S, District heating in sequential energy supply. Applied Energy 2012; 95: 123-131. [6] Joelsson A, Gustavsson L, District heating and energy efficiency in detached houses of different size and construction. Applied Energy 2009; 86: 126-134. [7] Åberg M, Henning D, Optimisation of Swedish district heating system with reduced heat demand due to energy efficiency measure in residential buildings. Energy Policy 2011; 39 7839-7852. [8] Lund H et al, The role of district heating in future renewable energy systems. Energy 2010; 35: 1381-1390. [9] Persson U, Werner S, Heat distribution and future competitiveness of district heating. Applied Energy 2011; 88:568-576. [10] Lücking G, Demand side Management in der Fernwärmeversorgung. Fernwärme international 1995 24; 6:286-294. 28

[11] Sperling K, Möller B, End-use energy savings and district heating expansion in a local renewable energy system A short-term perspective. Applied Energy 2012; 92: 831-842. [12] Wei B, Wang S-L, Li L, Fuzzy comprehencive evaluation of district heating systems. Energy Policy 2010; 38: 5947-5955. [13] Verda V, Coella F, Primary energy savings through thermal storage in district heating networks. Energy 2011; 36: 4278-4286. [14] Gustavsson S I, Rönnqvist M, Optimal heating of large block of flats. Energy and Buildings 2008; 40: 1699-1708. [15] Pengfei J et al, Modeling the dynamic characteristics of a district heating network. Energy 2012; 39: 126-134. [16] Dobos L, Abonyi J, Controller tuning of district heating networks using experiment design techniques. Energy 2011; 36: 4633-6439. [17] Steer K C B et al, Control period selection for improved operating performance in district heating networks. Energy and Buildings 2011; 43: 605-613. [18] Yliniemi K, Fault detection in district heating substations. Doctoral thesis, EISLAB, Department of Computer Science and Electrical Engineering, Luleå University of Technology, Luleå, 2005. [19] Gustafsson J, Distributed wireless control strategies for district heating substations. Licentiate thesis, EISLAB, Department of Computer and Electrical Engineering, Luleå University of Technology, Luleå, 2009. [20] Aronsson S, Fjärrvärmekunders värme och effektbehov (Heat and heat power demands for district heating customers). Doctoral Thesis, Department of Building Service Engineering, Chalmers University of Technology, Gothenburg, 1996. [21] Bøhm B, Danig P O, Monitoring he energy consumption in a district heated apartment building in Copenhagen, with specific interest in the thermodynamic performance. Energy and buildings 2004;.36: 229-236. 29

[22] Winberg A, Werner s, Avkylning av fjärrvärmevatten I befintliga abonnentcentraler.(district heating water chilling in substations), Stiftelsen för värmeteknisk forskning, Stockholm, 1987. [23] Danish district heating association, Udvikling av styring for regulering af omløb hos kunden (Development of strategy for by-pass control at substations). Dansk fjernvarmes F&U-Konto, Projekt nr. 2012-05, Kolding, 2012. [24] Wallin F, New opportunities provided by the Swedish electric meter reform. Doctoral Thesis, Mälardalen University, Västerås 2010. [25] Ratti C, Townsend A, The social Nexus. Scientific American, Volume 305, Number 3, September, 2011. [26] Euroheat & Power, Final Report Good Practice in Metering and Billing.. Available at: http://www.euroheat.org/reports/studies-27.aspx (2012-06-10) [27] Frederiksen S, Werner S, Fjärrvärme Teori, teknik och function (District heating Theory, technology and function). Studentlitteratur, Lund 1993. [28] 1999:716 Förordning om mätning, beräkning och rapportering av överförd el (Decree on metering, calculation and reporting of transporter of electricity). Swedish Ministry of Enterprise, 1999. [29] European Union, Directive 2006/32/EC of the European Parliament and council on energy end-user efficiency and energy service and repealing Council Directive 93/76/EEC. 2006. [30] Jomni Yassin, Improving Heat Measurement Accuracy in District Heating Substations. Doctoral Thesis, Department of Computer Science and Electronical Engineering, Luleå University of Technology, Luleå, 2006. [31] Swedish district heating association, Värmemätare, Tekniska branschkrav och råd om mätarhantering, Tekniska bestämmelser F104 (Technical requirements of Heat meter handling). Stockholm, 2008. 30

[32] Arvastson L, Stochastic modelling and operational optimization in district heating systems. Doctoral thesis, Centre for Mathematical Sciences, Lund institute of Technology, Lund, 2001. [33] Eriksson H, Short term Operation of District heating Systems. Doctoral thesis, Department of Energy Conversion, Chalmers University of Technology, Gothenburg, 1994. [34] Stevanovic V, et al, Prediction of thermal transients in district heating systems. Energy Conversion Management 2009; 50: 2167-2173. [35] Dotzauer E, Simple model for prediction of loads in district heating systems. Applied Energy 2002; 73: 277-284. [36] Nilsen, J R, Two-step decision and optimisation model for centralised or decentralised thermal storage in DH&C systems. IEA DHC report AnnexVII, 2005:8DHC-05.02, Borås, 2005. [37] Wernstedt F, Multi-agent systems for distributed control of district heating systems. Doctoral thesis, Department of Systems and Software Engineering, Blekinge Institute of Thechnology, Karlskrona, 2005. [38] Johansson C, Towards intelligent district heating. Licentiate dissertation, School of computing, Blekinge Institute of Technology, Karlskrona, 2010. [39] Wigbels M, Dynamic heat storage optimisation and demand side management. IEA DHC report Annex VII, 2005:8DHC-05.06, Oberhausen, 2005. [40] Olsson L, Werner S, Building mass used as short term storage. The 11 th International Symposium on District Heating and Cooling, Reykavik, 2008. 31

Paper I

Daily heat load variations in Swedish district heating systems Henrik Gadd 1 Sven Werner School of Business and Engineering Halmstad University, PO Box 823, SE-30118 Halmstad, Sweden Abstract Heat load variations in district heating systems have both seasonal and daily explanations. Seasonal variations are mainly caused by the seasonal variation in outdoor temperature, while daily variations are mainly induced from social patterns due to customer behaviours. Seasonal variations are well documented and analysed, but information about daily heat load variations is scarce. Heat storages can be used for elimination of daily heat load variations in the heat supply. In order to design these heat storages properly, the daily heat load variations must be quantified. The daily heat load variations have in this work been quantified by three defined key parameters. These parameters consider the annual relative heat load variation, the heat storage volume to counteract the load variations, and the loading and unloading capacity for this heat storage. These three parameters have then been estimated for time series of hourly average heat loads from 20 Swedish district heating systems, representing the whole range from small to large systems. The results show that the hourly heat load additions beyond the daily averages correspond to between 3% and 6% of the annual volume of heat supplied to the networks. Hereby, the daily variations are smaller than the seasonal variations, since the daily heat load additions beyond the annual average heat load are between 17% and 28 % of the annual volume of heat supplied to the networks. The size of short term heat storage to eliminate the daily heat load variations have been estimated to a heat volume corresponding to about 17% of the average daily heat supply into the network. This conclusion can also be expressed as a demand of 2.5 m 3 of heat storage volume per TJ of heat supplied, if assuming heat storage in water with a temperature difference of 40ºC. The capacity for loading and unloading the heat storage should be equal to about half of the annual average heat load for heat supplied into the network. Keywords District heating, daily variations, heat storage, heat load variation, seasonal variation 1. 1 Corresponding author: henrik.gadd@hh.se +46 35 167757 1

1 Introduction Heat deliveries in Swedish district heating systems are mainly used for space heating and domestic hot water preparation. Some industrial applications exist, but in many cases, the supply temperature in the district heating systems are too low to be used in industrial processes. Heat load in district heating systems is the aggregated heat load from the heat customers connected to the district heating network and the distribution losses. The heat supply is controlled by four independent factors: The first is the hot water taps and valves in radiators and ventilation air heating systems which control the heat demand. The second is the control valves in the primary side in the substation which keeps constant temperature of hot water and supply temperature to heating systems depending of outdoor temperature by controlling the primary flow. The third is differential pressure control on the primary side where the differential pressure has to be kept at a set point at the periphery of the network. The fourth is the supply temperature on the primary side depending in outdoor temperature. I.e. it is the heat users that are in control of the heat demand. The district heating operators deliver a possibility for a proper heat supply. Since the heat load at the customers is not constant, heat load variation at the customers result in heat load variation in the heat plant. The heat demand from a district heating system is fulfilled by a water mass flow and a temperature difference. I.e. there are two ways to satisfy changes in heat demand: Changing flow through all district heating sub-stations or changing temperature difference between supply and return pipe. If a customer increase the heat demand by increasing the mass flow, the increased heat demand propagate to the heat plant by the speed of sound in water i.e. approximately 1 000 m/s. But, if the customers increase the heat demand by increasing the temperature difference, the heat demand propagates to the heat plant with the flow rate of the water in the district heating pipes, i.e. 1 to 3 m/s. Hence, changes in heat demand due to changes in flow rate propagate to the heat plant in a period of a few seconds, while heat demand due to changes in temperature difference will propagate to the heat plant in minutes for the customers close to the heat plant and in hours for customers at the periphery of the district heating network, at least in large district heating systems. This is called geographical diversity. For further information about district heating system functions, see [1] Large variation in the outdoor temperature between summer and winter generate large heat load variations over the year, seasonal heat load variations, but there are also heat load variations between, and within single days, daily heat load variations. The aim of this paper is to develop a method to characterise and quantify daily heat load variations in district heating systems. In the literature several works can be found were suggestions to predict or control the heat load in the heat plant. In a model of heat load forecasting it is stated that especially the fast changes in heat load is difficult to predict [2]. Heat load prediction would make it possible to take action in advance. A prediction by simulating a repetitive heat load pattern is presented in [3] and [4]. A support for actions in the district heating network is describes in [5], where a method to predict how a temperature front propagates in a district heating network. By using multi-agent systems, were the substations and heat plant can communicate with one another, a possibility to control each part of the system, including the substations, and optimise the whole system would be possible. [6], [7] One possibility with this method would be if there are deficit of heat, the existing heat could be supplied to all heat customers instead of only the customer closest to the heat plant. A second possibility would be to use buildings as heat storage as described in [8] and thereby be able to use the entire district heating system, including the connected buildings, as heat storage. Advantages of daily heat load variation elimination are described in [9]. The possibility to optimise and reduce peak loads i.e. decrease daily heat 2

load variation is described in [10] by using heat meter measures at the customers as input information. It is obvious that daily variations can cause load problems. Various actions are taken in order to decrease daily heat load variations, but still it can not be found in the literature a method to quantify the daily heat load variations in district heating systems. More knowledge about heat load variations is needed in order to create smarter heat grids in the future. 1.1 Seasonal heat load variation Seasonal heat load variation is well known and obvious. It mainly depends on large difference in outdoor temperature between winter and summer, combined with the demand to have a more or less constant temperature inside the building envelopes. Heat loads can be split in two categories: physical heat load and social heat load. Heat loads that depends on physical conditions, like temperature difference and degree of insulation, is called physical heat load. Distribution losses are also physical losses since they depend in the temperature difference between the district heating water and the surrounding temperature of the district heating pipes. Other physical heat loads is influence of wind and solar radiation. Wind increases the heat demand because of infiltration. Warm air is replaced with cold air that has to be heated. Solar radiation decreases the demand of external heat in two ways. It increase the temperature of the, by solar radiation, exposed outer walls and thereby decrease the flow of heat from the inside of the building through the wall and windows acting like green house were solar radiation is let in to the building but the long wave radiation cannot pass throughout the window glasses. Both wind and solar radiation increases the seasonal heat load variations. The windiest parts of the year is when it is cold outside and the solar radiation is most intensive during the warm parts of the year. Social heat load depends in social behaviour of tenants. A typical social heat load is domestic hot water preparation. This preparation is an important factor for daily heat load variation as will be described further down. There is a seasonally component in hot water preparation as well. In the winter, people spend more time indoors and thereby use more hot water. In the summer and during holidays, tenants leave their urban dwellings temporarily and do not use hot water at all. Hence, domestic hot water preparation increases the seasonal heat load variation. A physical heat load part in domestic hot water is that the temperature of the incoming water is changing during the year. Especially in cities were fresh water is taken from a surface water reservoir. In this case, the seasonal heat load variation will increase since the incoming cold water is colder in winter than in summer. Further description of seasonal heat load variations in district heating system can be found in [11]. In Figure 1, a typical seasonal heat load pattern can be observed with high heat loads during winter and low heat loads during summer. 3

Daily average heat load [MW] 350 300 250 Annual average heat load (133 MW) 200 150 100 50 0 Figure 1. Seasonal heat load variation illustrated by the daily average heat load during a year in a district heating systems with an annual heat supply of about 4 400 TJ. 1.2 Daily heat load variation Heat demands in a district heating system are generated at the customers. These heat demands are not constant during the day. Even though district heating systems smoothen daily heat loads variations because of geographical diversity and that heat load peaks at the customers do not occur in the same moment, there are daily heat load variations. There are several reasons for daily heat load variations in district heating systems. Most of them are social heat demands. When a person choose to turn on a hot water tap to wash the hands it will result in an increased heat demand in the building that will reach the heat supply plant through the district heating network. Social heat demands are heat demands caused by both individual and collective social behaviours. One example of individual social behaviour is hot water use. Harmonised working times is an example of collective social behaviour. In offices and schools, where no people attend the buildings in nights and weekends, no or less ventilation rates can be applied. Hereby, time clock operation of ventilation should be used in all ventilated spaces that are not in use 24 hours a day. This action will decrease the heat demand, but it will also create daily heat load variations. In residential dwellings, tenants normally sleep at nights and do not use domestic hot water, but when they wake up, the first thing they do is to go to the bathroom and turn on the hot water tap. This behaviour will increase the heat demand and create daily heat load variations. The same thing will occur when people come home from work in the evening and start to use hot water. Night set back is still available in heating control systems even though it does not decrease the total heat use, but it increases the heat load variation in a way close to time clock operation of ventilation [12]. In domestic hot water preparation, two different methods are used: Direct hot water preparation and hot water storage. In direct preparation, the hot water is heated momentary at use in a heat exchanger having the capacity to fulfil all peak demands directly. The hot water 4

storage method has a heat exchanger with a lower capacity for loading the storage. At peak demands, the storage of domestic hot water is unloaded. When stored hot water is used, the hot water is replaced with cold water. When using direct preparation, the preparation coincides with the use, creating some daily heat load variations. When hot water storage is used, the daily load variations will not be so pronounced. There are also physical heat loads that generate daily heat load variations. The fact that night time outdoor temperatures are normally lower than day time temperatures generates daily heat load variations. Solar radiation also decreases the day time heat loads. A typical heat load pattern can be observed in Figure 2. Three characteristics can be identified: 1. Two peaks during a day, one heat power peak in the morning and one peak in the afternoon. 2. The influence of large differences in outdoor temperature during night and day in spring and autumn giving a significant dip in the heat load in the middle of the day. 3. No or small weekly heat load variations, i.e. variations between different days of the week. Average hourly heat load [MW] 350 300 250 200 December - February March-April & October-November 150 May & September 100 50 June - August 0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday Figure 2. Daily heat load variation illustrated by the aggregated average hourly heat load during weekdays for four different seasons in a district heating systems with an annual heat supply of about 4 400 TJ. 1.3 Consequences of heat load variation Heat load variations, both seasonal and daily, generate increased costs in district heating systems. Heat plants must always generate and supply the customer s aggregated heat power to the district heating network. A problem in district heating systems is that if not enough heat power is supplied to the network it does not affect all customers equally, but only the customer peripheral in the district heating network. If the heat supply to the district heating network is less than the heat demand, the customer closest to the heat plant will not notice any lack of heat, but customers at the periphery of the district heating network, will not get any or very little heat. The reason for this is that the differential pressure between supply and return pipe is highest close to the heat plant and then decreasing to the periphery of the district 5

heating network. At the customers, it is the differential pressure between supply and return pipe that drives the flow through the district heating substation. The differential pressure control is managed by the main distribution pumps in the district heating network and there are normally no local pumps at the customers. Therefore the customers closest to the heat plant will have the highest differential pressure and use it to increase the flow of district heating water through their district heating substations to increase the heat power while the customers at the periphery will suffer from lack of heat. As a consequence of this fact, most of the time an overcapacity of heat power has to be available to secure the heat supply to all heat customer. One way to handle the variation and the one that is mostly used is to have heat storages. In the 1980s, investigations and some tests were performed to have seasonal storages to store heat in the summer to be used in the winter, but so far no competitive technology has been found [13], [14]. The cost of the seasonal heat storage is too large compared to alternative heat cost. The only seasonal heat storage in operation known by the authors is located in Marstal, Denmark, where heat storage for a solar district heating system is in use.[15] For daily variations though, there are a number of possible methods to decrease peak load capacity and thereby decrease heat load capacity need in the district heating systems. Often it is a part of optimisation of heat storage where other influencing factors as increased electricity generation and maximising of industrial excess heat is included. Various examples of heat storage sizing are presented in [16] and [17]. It is stated in [17] that the optimal heat storage is strongly connected to the relative amount of relative peak load, i.e. how large the heat load variation is. The district heating network contains a large mass of water. By increasing the supply temperature above what is necessary for the present heat supply heat can be stored in the network. Another solution is to use heavy buildings connected to the district heating network as heat storages [8]. In heavy buildings with time constants of some hundred hours, it would be possible to increase the heat supply during low load in the district heating system and decrease the heat supply during high load without a reduction of customer heat comfort. If daily heat load variations could be eliminated in district heating systems, it would make the operation of the district heating system less costly and more competitive. There would be several advantages in the operation such as: Less use of expensive peak load power where often expensive fuels are used. Less need for peak load power capacity. Less need of electricity for district heating network pumping. Improved utilisation of industrial excess heat. Easier to optimize the operation that leads to higher conversion efficiencies. Less need for maintenance because of more smooth operation of the plants. Before water accumulators are built or devices installed to store heat in the district heating network or customers buildings, the heat load variations need to be characterized and quantified Even thou daily heat load variations is often mentioned as something that it desirable to eliminate no work has been found that closer describe nor quantify daily heat load variations. This paper presents a method to describe daily heat load variations in district heating systems by giving answers to these three questions: Which magnitude has the daily heat load variation in a district heating system? Which heat storage volume is needed in order to eliminate the daily heat load variations? 6

Which capacity is needed for loading and unloading this heat storage? 2 Method Time series of heat supplied into 20 Swedish district heating networks have been collected in order to analyse the daily heat load variations. The resolution in these time series is one hour, giving 8 760 hourly average heat loads values for one year for each system. In order to describe the daily heat load variations, three different variables have been defined: 1. Annual relative daily variation ( G a ). 2. Relative daily variation ( G d ). 3. Relative hourly variation ( G h ). To be able to compare daily variations with seasonal variations, a fourth variable for seasonal heat load variations has been defined: 4. Annual relative seasonal variation (D a ). The annual relative daily variation, G a, is a measure of the daily heat load variation during a year. The value itself express the annual proportion of the sum of all heat loads supplied over each daily average heat load over one year. It is used to compare different district heating systems with each other. The relative daily variation, G d, is a measure of the daily heat load variation for a single day. This variable expresses how much heat that needs to be stored each day in order to eliminate each daily heat load variation. A heat storage that store as much heat as the highest value of the year will give the possibility to eliminate all daily heat load variations. The sum of all relative daily variations divided by 365 becomes the annual relative daily variation. The relative hourly variation, G h, is the daily variation each hour. It expresses the heat transport capacity for loading and unloading the heat storage each hour. In order to be able to eliminate all daily heat load variations, a heat transport capacity to and from the heat storage equal to the highest relative hourly variation is needed. The annual relative seasonal variation, D a, is a measure of the seasonal heat load variation during a year. The value itself express the sum of all daily average heat loads supplied over the annual average heat load in one year. All four variables are relative variables related to the annual average heat load multiplied by number of hours related to the variable: 8 760 hours for the annual relative daily variation, and the annual relative seasonal variation, 24 hours for the relative daily variation, and 1 hour for the relative hourly variation. The four variables are determined from hourly average heat load (P h ), daily average heat load (P d ) and annual average heat load (P a ). 2.1 Annual relative daily variation ( G a) Annual relative daily variation is defined as: 7

8760,365 1 Ph Pd 2 h 1, d 1 G a 100 [%] (1) P 8760 where a P h =Hourly average heat load. P d =Daily average heat load. P =Annual average heat load. a [W] [W] [W] The annual relative variation is the accumulated positive difference between the hourly average heat loads and the daily average heat load during a year divided by the annual average heat load and the number of hours during one year. The division with the annual average heat load is introduced in order to get a measure independent of system size. The annual relative daily variation is expressed with one single value per system and year. The value itself express the annual proportion of all heat loads supplied over the daily average heat loads. These annual values can be used to compare the daily heat load variation from various district heating systems. 2.2 Relative daily variation ( G d ) Relative daily variation is defined as: 1 24 Ph Pd 2 h 1 G d 100 [%] (2) P 24 a The relative daily variation is the accumulated positive difference between the hourly average heat load and the daily average heat load divided by the annual average heat load and the number of hours during a day. The relative daily variation is expressed with 365 values per system and year. Relative daily variation is determined for each day and is a variable that quantify the amount of heat that divert from the daily average heat load. A heat storage size equal to the largest value of relative daily variation during a year is enough to eliminate all daily variations over the year. 2.3 Relative hourly variation ( G h) Relative hourly variation is defined as: Ph Pd G h 100 [%] (3) P a 8

The relative hourly variation is the absolute difference between the hourly average heat load and the daily average heat load divided by the annual average heat load. The relative hourly variation is expressed with 8 760 values per system and year. The relative hourly variation is the heat power capacity for loading and unloading the heat storage to eliminate the daily variations. A heat power capacity of loading and unloading the heat storage equal to the largest value of relative hourly variation is the amount of heat load capacity to eliminate daily heat load variations over the year. 2.4 Annual relative seasonal variation(d a ) The annual relative seasonal variation is defined as: 1 365 Pd Pa 2 d 1 D a 100 [%] (4) Pa 8760 The annual relative seasonal variation is the accumulated positive difference between the daily average heat loads and the annual average heat load during a year divided by the annual average heat load and the number of hours during one year. The division with the annual average heat load is introduced in order to get a measure independent of system size. The annual relative seasonal variation is expressed with one single value per system and year. The value itself express the annual proportion of all heat loads supplied over the annual average heat load. 9

3 Gathered Data The data sets that are used to determine the daily heat load variations have been collected from 20 district heating systems in Sweden. It is the heat supply to the district heating network, i.e. distribution losses are included in the measuring values. The sizes of the analysed district heating systems are between 32 TJ and 13 300 TJ heat supplied annually to the networks. The data sets is one year series from 1 st of January to 31 st of December, i.e. 8 760 values each year. The unit of the values from the meter reading system is MWh/h. Most of the data sets are from 2008 and 2009, but a few from the years 2004-2007. For most district heating systems one year data set is collected, but for two district heating systems 5 respectively 6 years data sets are collected. This multiyear data is used to analyse the daily heat load variation between different years, later presented in the result section. The hourly average heat load is often called heat power, but it is actually delivered energy during one hour. The heat is continuously measured every whole hour. The present meter value minus the preceding meter value is the hourly value for the present hour. District heating systems with an annual heat supply of more than 700 TJ are normally always measured hourly. For district heating networks with an annual heat supply less than 350 TJ, the number of hourly measured systems is fast decreasing. For most of the data series, no or single values are missing, but in a few cases, values in the annual data series are not complete. For single values and up to 5 values in a row missing values are reconstructed by interpolation. Since district heating network systems are thermally slow, changes in the heat power are also slow. If there are more than 5 values missing, an analysis of the day before and after was made to see if there is a typical heat load pattern. Sometimes it is possible to interpolate for more than 5 values. If the patterns does not show a linear behaviour the values from either the day before or the day after is copied. Which day that is used depends of which of the days that looks most like the day with the incomplete data series. The amounts of corrected values are less than one per thousand and have thereby no impact in the results. 10

4 Results 4.1 Annual relative daily variation As can be observed in Figure 3, the annual relative daily variation does not differ much between district heating systems. In the studied district heating systems, the annual relative daily variation in are between 5.7% and 2.6% of the annual volume of heat supplied into the district heating networks with a mean value of 4.5%. It is only two large systems that have somewhat lower annual relative daily variation. If the same method is used for seasonal heat load variation as for daily variation the annual relative seasonal variation is in average 24% with a spreading between 17-28%. The annual relative seasonal variation for system in Figure 1 is 27% which corresponds to the area over the mean value divided by the total annual heat supply. Compared to the seasonal heat load variation, daily heat load variation is very low. The reason for this is that the main part of heat demand is caused by the difference between outdoor and indoor temperature. Annual relative daily variation [%] 6% 5% 4% 3% 2% 1% 0% 10 100 1 000 10 000 100 000 Annual heat supply [TJ] Figure 3. Annual relative daily variation for the 20 Swedish district heating systems analysed. An expected result would be that large district heating systems have smaller relative daily variations (G a ) than small district heating systems. There are two reasons for that: 1. In large district heating networks, the customers is geographically spread on different pipe distances from the heat supply plants. Hereby, the water in the return pipe arrives to the heat supply plants at different times compared to when the return water left each substation. This is called geographical diversity in the district heating network and should reduce the daily heat load variation. 2. In large district heating networks, you would expect that the operators have more active operation in the heat distribution network with respect to temporary heat storage in the district heating network. 11

But as can be observed in Figure 3 it is not as simple as that. The daily heat load variation is more or less the same for most district heating systems. For district heating systems larger than 4 000 TJ annual delivered heat seem to have smaller daily heat load variations. The annual relative daily variation for the Swedish total electricity use 2008 was as a comparison 4.5 %. With a total delivered amount of electricity of 497 PJ [18]. I.e. about the same size of daily heat load variation as district heating systems. One district cooling system with 72 TJ of annual supplied cold has also been evaluated. The annual relative variation in this system was 8.6%, which is twice the daily variation estimated for the district heating networks. To investigate if the annual relative daily variation differs from one year to another, multi-year data series were gathered for two district heating systems. Both district heating systems are located in south of Sweden, one with an annual heat supply of 700 TJ and one with an annual heat supply of 8 000 TJ, Figure 4. The difference is 0.44 percentage units for the larger system and 0.65 percentage units for the smaller system. I.e. difference in annual relative daily variations between different years is small. Annual relative daily variation 10% 9% 8% 7% 6% 5% 4% 3% 2% 1% 0% 2003 2004 2005 2006 2007 2008 2009 2010 Year Annual heat supply 8 000 TJ Annual heat supply 700 TJ Figure 4. Annual relative daily variation across years for two Swedish district heating system with an annual heat supply of about 700 TJ respectively 8 000TJ. The multi-year data sets have been analysed to see if there are any correlation between annual heat supply and annual relative daily variation. No such correlation could be found. 4.2 Relative daily variations In Figure 5, the relative daily variation is calculated for each day during the year and sorted by magnitude for the 20 Swedish district heating systems analysed. Figure 6 contains the same information, but focusing on the highest values obtained. Maximum value for the largest daily heat load variation is 30% of daily average heat supply and minimum value is 9%. Average value is 17%. In absence of economical evaluation the 99 th percentile could be used as design condition to exclude extreme values. 12

For the 99 th percentile, corresponding to 3.65 days, the max value has decreased to 15%, min value to 8%, and average value to 12%. With an effective heat storage size corresponding to 12% of daily average heat load, almost all daily heat load variations are possible to eliminate. The conclusion above can also be used for estimating the specific demand for a heat storage volume. For each TJ of annual supplied heat, the annual average heat load becomes 32 kw. With 3.6 hours of operation (15% of 24 hours), this will give a demand for storing 410 MJ. Assuming 40 C temperature difference for the heat storage, the requested water volume becomes almost 2.5 m 3 for each TJ of heat supplied during a year. Corresponding heat storage volume for the relative daily variation of 12% is 2 m 3 /TJ annually heat supply. A brief study has been performed for existing heat storages in Swedish district heating systems. The result is that the sizes of heat storages installed are between 4% and 250% with an average of 47 % of average daily supplied heat. I.e. the average heat storage in Swedish district heating systems is 3 times larger than the heat storage size demand to eliminate daily heat load variations. This indicates that heat storages are used to do more than eliminate daily heat load variations. 13

Relative daily variation [%] 30% 25% Enlarged area in Figure 6 20% 15% 10% 5% 0% 0 50 100 150 200 250 300 350 Number of days Figure 5. Relative daily variation for the 20 Swedish district heating systems analysed. Relative daily variation [%] 20% 15% 10% 5% 0% 0 5 10 15 20 25 30 35 40 45 50 Number of days Figure 6. Enlarged part of Figure 5 showing the peak parts of the relative daily variation. 14

4.3 Relative hourly variation In Figure 7, the relative hourly variation is calculated for each hour over a year. Note that it is the absolute value, i.e. it can be either loading or unloading capacity. Figure 8 contain the same information but magnified for the high values. Highest value of all 20 district heating systems is 196 % of annual average heat load and the lowest is 46%. Average value is 91%. With the same reason as for relative daily variations, i.e. absence of economical evaluation, the 99 th percentile could be used as sizing condition to exclude extreme values. In the 99 th percentile the max value has decreased to 47%, min has decreased to 23% and the average value to 38%. On the analogy of relative daily variation a specific heat loading and unloading capacity for the heat storage to eliminate heat load variations can be quantified. An hourly relative variation of 38 % results in a loading/unloading capacity of 12 kw for each TJ supplied. Relative hourly variation [%] 150% 100% Enlarged in Figure 8 50% 0% 0 1 000 2 000 3 000 4 000 5 000 6 000 7 000 8 000 Number of hours Figure 7. Relative hourly variation for the 20 Swedish district heating systems analysed. 15

Relative hourly variation [%] 70% 60% 50% 40% 30% 20% 10% 0% 0 50 100 150 200 250 300 350 400 450 500 Number of hours Figure 8. Enlarged part of Figure 7 showing the peak parts of the relative hourly variation. 16

5 Conclusions The magnitude of annual relative daily variation is small compare to annual supplied heat and compared to seasonal heat load variations. While the average annual relative variation is 4.5 % for the evaluated 20 district heating systems the average annual relative seasonal variation is 24%. The size of heat storage to eliminate daily heat load variations is in the magnitude of 15% of average daily supplied heat. The size of existing heat storages installed in district heating systems in Sweden are in an average three times the necessary to eliminate daily heat load variations. This indicates that heat storages are used for more than eliminate daily heat load variations. Two examples could be storage as a reserve between days or to increase electrical power generation at high prices. Hourly heat load variation is just below 40% of the annual average heat load. This correspond to a loading and unloading speed of 7 hours from full to empty or empty to full heat storage. In order to better understand the daily variations in district heating systems, the corresponding variations should be analysed for the customer substations. A natural future work would be to study these local daily variations in order to identify if and how these variations can be eliminated. Otherwise, there is an obvious risk that heat storages are installed for a problem that can be solved by a lower cost at the variation source. 17

6 Acknowledgements This work was financially supported by Fjärrsyn, the Swedish district heating research programme, and Öresundskraft. The support from the twenty district heating companies providing input data for the analysis was also highly appreciated. 7 References [1] Frederiksen S, Werner S, Fjärrvärme, Teori, teknik och funktion (District heating Theory, Technology and Function). Studentlitteratur, Lund 1993. [2] Wiklund H, Short term forcasting of heat load in a district heating system. Fernwärme international, 20(1991) 286-294. [3] Arvaston L, Stochastic modelling and operational optimization in district heating system. Doctoral thesis, Mathematical Statistics, Lund University, Lund 2001. [4] Dotzauer E, Simple model for prediction of loads in district-heating systems. Applied Energy 73(2002) 277-284. [5] Stevanovic V D et al, Prediction of thermal transients in district heating systems. Energy Conversion and Management 50(2009) 2167-2173. [6] Wernstedt F, Multi-agent systems for distributed control of district heating systems. Doctoral thesis, Department of Systems and Software Engineering, School of Engineering, Blekinge Institute of Technology, Karlskrona 2005. [7] Johansson C, Towards intelligent district heating. Licentiate Dissertation, School of Computing, Blekinge Institute of Technology, Karlskrona 2010. [8] Olsson L, Werner S, Building mass used as short term heat storage. The 11th International Symposium on District Heating and Cooling, Reykjavik 2008. [9] Verda V, Colella F, Primary energy savings through thermal storage in district heating networks. Energy 36(2011) 4278-4286. [10] Drysdale A et al, Optimised district heating systems using remote heat meter communication and control. IEA DHC Annex VI Report 2002:S7, DTI Taastrup 2003. [11] Werner S, The heat load in district heating systems. Doctoral thesis, Department of Energy Conversion, Chalmers University of Technology, Gothenburg 1984. [12] Aronsson S, Fjärrvärmekunders värme och effektbehov (Heat and heat power demands for district heating customers). Doctoral Thesis Department of Building Service Engineering, Chalmers University of Technology, Gothenburg 1996. [13] Hydén H, Töcksberg B, Potentian för säsongslagring av värme i svenska fjärrvärmesystem (Potential for seasonal heat storage in Swedish district heating systems). Rapport R112:1985, Byggforkningsrådet,Stockholm, 1985. [14] Gabrielsson E, Seasonal storage of Thermal Energy-Swedish Experience. Fernwärme International, 1990; 19, nr 3: 221-234. [15] Marstal Fjärnvärme, Solar Thermal and Long Term Heat Storage for District Heating Systems. Final Technical report, NNE5-2000200490, 2005. [16] Wigbels M, Dynamic storage optimisation and demand side management. IEA DHC report Annex VII, 2005:8DHC-05.06. Oberhausen 2005. [17] Nielsen J R, Two-step decision and optimisation model for centralised or decentralised thermal storage in DH&C systems. IEA DHC report Annex VII, 2005:8DHC-05.02, Borås 2005. 18

[18] Svenska Kraftnät, Statistik för Sverige 2008 (Swedish national grid, Statistics for Sweden 2008). downloaded from:http: www.svk.se/global/06_energimarknaden/xls/statistik/n_fot2008.xls, [2010-06- 15. 19

Paper II

Heat load patterns in district heating substations Henrik Gadd 1 Sven Werner School of Business and Engineering Halmstad University, PO Box 823, SE-30118 Halmstad, Sweden Abstract Future smart energy grids will require more information exchange between interfaces in the energy systems. One interface where dearth of information exists is in district heating substations. The substations are the interface between the distribution network and the customer building heating system. Previously meter reading data was manually collected once or a few times a year. Today, automatic meter reading systems is being built up resulting in hourly meter reading data are available at low cost. The main purpose with this article is to perform an introductory analysis of high resolution measurements in order to provide valuable information about district heating substations for future use in smart heat grids. However, the heat load in a district heating network is the aggregated heat load from all customer substations connected to the network. Errors and deviations in customer substations will then propagate through the district heating network to the heat supply plants. In order to reduce future customer and heat supplier costs, a demand has appeared for smart functions identifying errors and deviations in customer substations. Hereby, a research demand appears for defining normal and abnormal heat load patterns in customer substations. Hourly heat meter readings from customer substations are nowadays available from automatic meter reading systems, giving the possibility to analyse customer heat loads from many various substations with a high time resolution. In this project, one year of heat meter readings from 141 district heating substations in two district heating networks were analysed. The connected customer buildings were classified into five different customer categories and four typical heat load patterns were identified. Two descriptive parameters, Annual relative daily variations and Annual relative seasonal variations, were defined from each one year sequence for identifying normal and abnormal heat load patterns. The three major conclusions are associated both with the method used and the objects analysed. Firstly, normal heat load patterns vary with applied control strategy, season, and customer category. Secondly, it is possible to identify obvious outliers compared to normal heat loads with the two descriptive parameters used in this initial analysis. Thirdly, the developed method can probably be enhanced by redefining the customer categories by their indoor activities. 1. 1 Corresponding author: henrik.gadd@hh.se +46 35 167757 1

Key words: District heating, heat load variation, automatic meter reading, heat load pattern, smart heat grids, smart energy grids 2

1. Introduction Future smart energy grids will require more information about the energy flows in various interfaces in the energy system according to [1].This information is not always available today for most interfaces. One interface where dearth of information exists is substations in district heating systems. These substations constitute the interface between the distribution network and the customer building heating system. This existing dearth of information can be explained by the previous lack of measurements, since large amount of data required to perform these analysis have not, by reasonable cost, been possible to collect. Previously meter reading were performed manually once or a few times a year. However, automatic meter reading systems are now being installed which makes hour meter readings available at low cost. The main purpose with this article is to perform an introductory analysis of high resolution measurements in order to provide valuable information about district heating substations for future use in smart heat grids. This is a novel area of research and there by noting is published in international journals before. In the past, efforts have been put in to optimise the operation of heat supply plants and district heating networks and to discover and eliminate corresponding errors and deviations. Heat load patterns from customer substations have often been taken for granted, both in design and in operation. However, the heat load in a district heating network is the aggregated heat load from all customer substations connected to the network. Errors and deviations in customer substations and internal heating systems in buildings will then propagate through the district heating network to the heat supply plants. In order to reduce future customer and heat supplier costs, a demand has appeared for more intelligent functions identifying errors and deviations in customer substations and heat supply systems in connected buildings. Hereby, a research demand appears for defining normal and abnormal heat load patterns in customer substations. The operation of the heating and ventilation systems in a building is shifting depending on the activity in the building. In schools, where no or few people are present during nights and at weekends, no or little ventilation is necessary at these times. During school holidays, the temperature in school buildings can be reduced. But multi-dwelling buildings need to be heated and ventilated 24 hours a day, 7 days a week, all year round. Hence, the heat load pattern is different from building to building depending on what kind of activity that takes place in the building. The best would of course be to make sure that the customers facilities are working well, but with hundreds or thousands of customer substations, it has until now been economically impossible to monitor all customer substations. Today, with automatic meter reading systems installed in most district heating systems in Sweden, new opportunities arise to systematically identify errors in the heat supply or control settings at the customers. If an error in a customer substation can be identified and eliminated, it may of course lead to less heat being sold, but the risk is that if it is not eliminated, the company may lose the total heat sales to the customer depending on the fact that other heating alternatives can be more competitive. Very few studies have been performed concerning horizontal analyses of the heat load pattern in a large number of substations. The reason is that before the large amount of data required to perform these analysis have not, by reasonable cost, been possible to collect. Automatic meter reading systems now installed makes hour meter readings available at low cost.. One work where heat load patterns have been analysed for 50 buildings is [2], where the main aim was to estimate heat load capacities for billing purposes. In order to increase energy efficiency in multi dwelling buildings, heat loads has been monitored and evaluated in [3]. 3

There are works about indoor comfort like [4] were thermal inertia in a building is evaluated which indirect is about heat load patterns. Characteristic for [2], [3] and [4] is that specific equipment have to be installed in buildings in order to collect measuring data which was, before meter reading systems was installed, what was available. A method of error detection in district heating substations by using information from billing systems is presented in [5]. There are studies performed in order to optimise the substation, often with the goal to decrease the primary return temperature as in [6], [7], [8] and [9]. There is also a study to identify faults in substations where a method to identify temperature sensor fault is described [10]. In this study there is also a method described for separating hot water use from space heating which from a heat load pattern point of view is very interesting This introduction forms a background to answer three research questions in a field of research which in many ways is a white spot on the district heating knowledge map: How do heat load patterns vary in substations? Can heat load patterns be simplified to identify outliers by using heat meter readings? In what plausible directions can this early research on substation heat load be enhanced? 2. Method Heat load patterns are not the same in all buildings. It depends on the building properties but also of the type of activity that takes place in the buildings. To be able to evaluate if a heat load in a building is normal or not, it is necessary to know what heat load pattern is to be expected. From the customer records at one district heating company, 141 buildings have been selected to be analysed. In the company customer records, seven customer categories are available of which five are used in this study. Two descriptive parameters and four heat load patterns are identified for each data set and plotted in diagrams presented in the results section. 2.1. Gathered Data The collected data sets are meter readings from 141 buildings connected to the district heating systems in Helsingborg and Ängelholm in the south-west of Sweden. In total, there are about 13 000 buildings connected to the two district heating systems from which about 10 000 are one- and two-dwelling buildings. The data sets are hourly measured one-year series from 1st of January to 31st of December, i.e. 8760 values annually for each building. All data sets are from the year 2010. The metering data sets come from databases in the automatic meter reading systems. In a few cases, single unreasonable one-hour-values appear in the data sets. They have been corrected by interpolation from the surrounding values. The unit of the values from the meter reading system is kwh/h. The values are often called heat powers, but it is actually delivered heat during one hour. They could also be referred to as hourly average heat loads. 2.2. Customer categories In the company customer records, the customer buildings are split into different types of customer categories depending on the activity in the buildings. The subdivision is made due to governmental demands to report statistical data that is collected each year. The national categories for customer categories in the national district heating statistics are: Manufacturing 4

industries, one- and two-dwelling buildings, multi-dwelling buildings, ground heating, public administration, and others. In this study one- and two-dwelling buildings and ground heating have been excluded. The reason for excluding one- and two-dwelling buildings is that they use less heat per building. It takes the same effort to eliminate a fault in a small building as in a large building, but there is probably less to gain. Ground heating deliveries differ from other usage of district heating since it is the heat in the return pipe that is used in the application and only less than 0.5% of the district heating deliveries in Sweden are supplied for ground heating purposes [11]. In the company customer records for the used heat meter data, the subdivision in different categories has in some cases a higher resolution. The main part of the buildings in the group Others in the national statistics is in the company customer records sorted under the category Commercial buildings. Public administration from the national statistics is split into Public administration and Health and Social Services. In this study, the analysis is split into the following five different customer categories: Multi-dwelling buildings Industrial demands Health and Social Services buildings Commercial buildings Public administration buildings 2.3. Two descriptive parameters In this paper, two descriptive parameters determined from heat energy metering values will be evaluated for different customer categories: Annual relative daily variation and Annual relative seasonal variation. Annual relative daily variation is a variation in the heat load compared to the daily mean heat load and is defined and described in [12]. Annual relative daily variations occur mainly because of social heat loads such as domestic hot water preparation and time clock operation control of ventilation, but also some physical heat loads that generate daily variation such as wind, solar radiation and temperature variations between night and day. Annual relative seasonal variation is the consequence of large variations in outdoor temperature between winter and summer, while the temperature inside buildings is expected to be constant. The first descriptive parameter, Annual relative daily variation, is defined as: 8760,365 1 Ph Pd 2 h 1, d 1 G a 100 [%] (1) P 8760 a where P h = Hourly average heat load P d = Daily average heat load P a = Annual average heat load [W] [W] [W] The annual relative daily variation is the accumulated positive difference between the hourly average heat loads and the daily average heat load during a year divided by the annual average heat load and the number of hours during one year. The division with the annual average heat load is introduced in order to get a measure independent of building size. The second descriptive parameter, Annual relative seasonal variation, is defined as: 5

365 1 24 2 P d Pa d 1 W 100 [%] (2) Pa 8760 The annual relative seasonal variation is the accumulated positive difference between the daily average heat loads and the annual average heat load during a year multiplied by the number of hours in one day and divided by the annual average heat load and the number of hours during one year. As for annual relative daily variation, the division with the annual average heat load is introduced in order to get a measure independent of each heat demand. 2.4. Heat load patterns Different types of buildings have different heat load patterns depending on the activity in the building. The heat load pattern changes during the year because of changing outdoor temperature levels during the year. For this reason, each one year sequence meter data set is split into four different season periods: Winter: December, January, February (average hourly values from 12 or 13 weekhour values) Early spring, late autumn: March, April, October, November (average hourly values from 17 or 18 week-hour values) Late spring, early autumn: May, September (average hourly values from 8 or 9 week-hour values) Summer: June, July, August (average hourly values from 13 or 14 week-hour values) For each period the average value is for every hour during a week, where Monday 00.00-01.00 is the first hour and Sunday 23.00-24.00 is the last in each week, plotted in a diagram. One diagram for each building has been plotted. The result is a weekly heat load pattern. Since it is an average value for between 8 to 18 values only recurrent heat load behaviours will appear. From the heat load pattern diagrams four different heat load patterns have been manually identified: Continuous operation control, Night setback control, Time clock operation control 5 days a week and Time clock operation control 7 days a week, which are described below. These heat load patterns have not been verified by substation visits or inspections of control routines. 6

2.4.1. Continuous operation control No additional control is applied other than keeping the indoor temperature at the set point in the building heating control system. For a well-insulated and not too small building, it will mainly be domestic hot water preparation that causes the heat load variations in the hourly time scale. Ventilation is in operation 24 hours a day. This is the typical control situation for residential buildings and some Health and Social Services buildings. A typical heat load pattern for continuous operation control can be observed in Figure 1. Small differences in heat load appear especially in winter and summer. In autumn and spring, reduced daytime heat loads can be observed. These are the results of additional heat contributions from solar irradiation to space heating. Average hourly heat load [kw] 600 500 December - February 400 300 200 March-April & October-November May & September 100 June - August 0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday Figure 1 Average weekly heat load patterns for continuous operation control during four season periods: Multi dwelling buildings with an annual heat supply of 2484 MWh or 8940 GJ. 7

2.4.2. Night setback control Night set back control is when the set point for the indoor temperature is lowered during the night. The traditional thought behind this control strategy is to get a lower indoor temperature during nights and thereby decrease the total heat demand. But most buildings have nowadays high time constants, giving a slow reduction of the indoor temperature due to appropriate insulation and airtight building envelopes. The indoor temperature will not decrease so much that a noticeable heat demand reduction will occur. The only result of night set back applied to energy efficient buildings is to move some heat load from nights to mornings. Hence, night setback control is only suitable and profitable for buildings with high specific demands and short time constants due to bad insulation and non-airtight building envelopes. A typical heat load pattern for night setback control can be observed in Figure 2. Lower heat loads during nights are followed by high peak heat loads in the mornings, but these peaks vanish quite fast. The peaks are the results of the reheating of the cooled off heating system during the preceding nights. Average hourly heat load [kw] 200 180 December - February 160 140 120 March-April & October-November 100 80 May & September 60 40 20 June - August 0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday Figure 2 Average weekly heat load patterns for night setback control during four season periods: Public administration building with an annual heat supply of 583 MWh or 2100 GJ. 8

2.4.3. Time clock operation control 5 days a week Ventilation in a building does not necessarily have to be in operation 24 hours a day 7 days a week. Schools, for example, only have daytime activities from Mondays to Fridays. At nights and weekends, no or few people are in the buildings and no or reduced ventilation will be appropriate. Full operation of the ventilation systems just increases the amount of used heat energy for the customer. For working days activities only, time clock operation control can be applied 5 days a week. A typical heat load pattern for time clock operation control 5 days a week can be observed in Figure 3. Note that the heat load during nights and weekends is the same. During these periods the ventilation is turned off or reduced and the radiator system is supplying heat to keep the indoor temperature at a desirable level. Average hourly heat load [kw] 180 160 December - February 140 120 100 80 60 40 20 March-April & October-November May & September June - August 0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday Figure 3 Average weekly heat load patterns for time clock operation control 5 days a week during four season periods: Public administration building with an annual heat supply of 432 MWh or 1560 GJ. 9

2.4.4. Time clock operation control 7 days a week Some buildings have a daytime use 7 days a week. One example is a shopping mall that is open 7 days a week in daytime. Still the ventilation can be shut off during the night since no or few people are inside the building at these times. A typical heat load pattern for time clock operation control 7 days a week can be observed in Figure 4. The pattern is similar to time clock operation control 5 days a week, but the ventilation is also in operation at the weekends as well and not only during working days. Average hourly heat load [kw] 450 400 December - February 350 300 250 200 150 100 50 March-April & October-November May & September June - August 0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday Figure 4 Average weekly heat load patterns for time clock operation control 7 days a week during four season periods: Commercial building with an annual heat supply of 1246 MWh or 4490 GJ. 10

3. Results The relative seasonal variation for heat loads in buildings is most dependent on customer category, and the type of activity in the buildings. Industrial, commercial, and public administration buildings have a relative seasonal variation of around 30 to 40%, independent of the annual relative daily variation. Health and Social Services buildings have around 30% and multi-dwelling buildings have the lowest relative seasonal variation between 20 and 30%. The annual relative daily variation has a large range in industrial, commercial and public administration buildings. Since most of these buildings should have time clock operation control of ventilation, they should also have large annual relative daily variations. The results in Figure 5 to Figure 9 indicate that time clock operation control of ventilation generates high annual relative daily variations. Still, there are in every group of building types some that seem to have too low or too high annual relative daily variation. Notable are the outliers that deviate from what seems to be a normal heat load pattern. 11

3.1. Multi-dwelling buildings Multi-dwelling buildings are relatively homogeneous types of buildings with respect to heat load patterns. They are in use 24 hours a day all year around and domestic hot water share of the heat load is relatively high, about 20% of the annual heat demand according to [13]. Multi-dwelling buildings are characterised by low annual relative daily variation. As can be seen in Figure 5, most of the multi-dwelling buildings have heat load patterns from continuous operation control. Only a few buildings seem to have some kind of night setback. Typical values for annual relative daily variation are between 4 and 8%. The relative seasonal variation is in the upper range compared to the other types of buildings in this study. The multi-dwelling buildings have an annual relative seasonal variation in the range of 22 to 32%. Most of the buildings are well gathered in the diagram, but there are 4 outliers. Most notable is the building with night setback heat load pattern with 24% annual relative daily variation but also the buildings with low annual relative seasonal variations are notable. It indicates a low correlation between heat load and outdoor temperature. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 10% 20% 30% 40% 50% Annual relative seasonal variation Continuous Night setback Figure 5 Annual relative daily variation as a function of Annual relative seasonal variation for 37 multi-dwelling buildings. 12

3.2. Health and Social Services buildings Health and Social Services buildings can be anything from a hospital to an office for the administrators and are thereby a very heterogeneous group. Some buildings like hospitals have a heat load pattern close to multi-dwelling buildings with 24 hour activity every day. Other buildings have just daytime activities and have heat load patterns close to traditional office buildings with time clock operation control of ventilation and low domestic hot water use. Remarks in Figure 6 are as follows: one building with a heat load pattern from continuous operation control and an annual relative daily variation of 13%, and one building with a heat load pattern from time clock operation control 7 days a week, but only 10% of annual relative daily variation. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 10% 20% 30% 40% 50% Annual relative seasonal variation Continuous Night setback Time clock operation 5 Time clock operation 7 Figure 6 Annual relative daily variation as a function of Annual relative seasonal variation for 11 Health and Social Services buildings. 13

3.3. Industrial demands The definition of industrial buildings is that they are used for the manufacture of materials or products. Their heat demands are more diversified than multi-dwelling buildings. There can be between one- to five-shift operations and thereby everything between 8 to 24 hours per day of activity. Heat demands can appear for both space heating and industrial processes. Some industries have excess heat and can thereby decrease their external heat demands partly. In most industrial buildings, there is no or less activity during nights and weekends which is why time clock operation control of ventilation is appropriate. Domestic hot water use is normally low compared to multi-dwelling buildings, i.e. summer heat load when no space heating is required ought to be low. As can be observed in Figure 7 fewer than half of the industrial customers seem to have time clock operation control of ventilation. A large portion of continuous heat load pattern indicates that the ventilation or other heat demands are running 24 hours a day in lots of industrial buildings. Most notable is the building with 4% annual relative seasonal variation and 2% annual relative variation. It is a more or less constant heat load over the year. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 10% 20% 30% 40% 50% Annual relative seasonal variation Continuous 14 Night setback Time clock operation 5 Time clock operation 7 Figure 7 Annual relative daily variation as a function of Annual relative seasonal variation for 36 industrial customers.

3.4. Commercial buildings Few commercial buildings are in operation at night. This is confirmed by the fact that most commercial buildings have heat load patterns from time clock operation control during 5 or 7 days a week. Still, there are some customers with a heat load pattern from continuous operation control. Commercial buildings consist of trading companies, restaurants, hotels, service companies, amusement and recreational services. These are buildings where activities take place mainly during the daytime 5 to 7 days a week. These buildings should have time clock operation control of ventilation. The use of domestic hot water is low. An exception is hotels that have a heat load pattern close to multi-dwelling buildings with 24 hour operation and a rather high share of domestic hot water of the heat load. Notable buildings in Figure 8 are three buildings with a heat load pattern of continuous operation control, but with relatively high annual relative daily variations. There are also three buildings with heat load patterns of time clock operation control during 7 days a week with notably low annual relative daily variation. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 10% 20% 30% 40% 50% Annual relative seasonal variation Continuous Time clock operation 5 Time clock operation 7 Figure 8 Annual relative daily variation as a function of Annual relative seasonal variation for 22 commercial buildings. 15

3.5. Public administration buildings Typical public administration buildings are schools and municipal administration buildings, that are mainly in use during office hours five days a week, gymnasiums, public baths, that are also used at weekends, but also fire stations and police stations with a 24 hour operation. In Figure 9 this is confirmed by heat load patterns from Continuous operation control, Time clock operation control 5 days a week and Time clock operation control 7 days a week. The use of domestic hot water is shifting, but it is low compared to multi-dwelling buildings. Three buildings are noteworthy with low annual relative seasonal variation. Also one building with a continuous heat load pattern of 15% annual relative daily variation is notable. Annual relative daily variation 25% 20% 15% 10% 5% 0% 0% 10% 20% 30% 40% 50% Continuous Annual relative seasonal variation Night setback Time clock operation 5 Time clock operation 7 Figure 9 Annual relative daily variation as a function of Annual relative seasonal variation for 35 public administration buildings. 16

3.6. Cross-cutting results The different types of buildings can be put into three different larger groups depending on variation in annual relative daily variation. Low annual relative daily variations: Multi-dwelling buildings Intermediate annual relative daily variations: Health and Social Services buildings High annual relative daily variations: Commercial, Industrial, and Public administration customers. The most important cause for high annual relative daily variation is time clock operation control of ventilation. In buildings with activity only parts of the day or week, ventilation is reduced or shut off when no indoor activities take place. In an office, normally no or very few people are in the building at nights and weekends. In a multi-dwelling building though, tenants are using heat 24 hours a day all year around. Another setting that increases annual relative daily variation is night setback control. Even though, night setback control does not have an influence on heat demand reduction, it is still not unusual that night setback controls are applied. A heavy building with a thermal time constant of at least 100 hours, which is the case with all the buildings in this study, will not cool off during a few night hours. The only results are large heat load peaks when the set point for the indoor temperature changes.the only thing that cools off is the ventilation and heating system, and in the morning, when the set point changes, a high heat load peak is a consequence to warm up the heating and ventilation system. To enhance the method in this paper, an inventory of the buildings to confirm the settings for Continuous operation control, Time clock operation control 5 days a week, Time clock operation control 7 days a week and Night setback control should be performed. This inventory together with a more suitable subdivision of customer categories that merge with an expected heat load pattern would increase the resolution of the method. It could either be a finer subdivision of the existing customer categories or an entirely new subdivision. The result could be used as an input to develop a method to automatically identify heat load patterns and thereby identify outliers. 4. Conclusions The three major conclusions are associated both to the method used and the objects analysed. Firstly, normal heat load patterns vary with applied control strategy, season, and customer category. High annual relative daily variation in a multi-dwelling building would indicate that something is wrong, but on the contrary, on commercial premises and in industries there is something wrong, if there are not high annual relative daily variations. But as can be observed in the results section, it is not an unambiguous result. A large variation of heat load patterns among various buildings implies that a standard heat load pattern for customer substations does not exist. Secondly, it is possible to identify obvious outliers compared to normal heat loads with the two descriptive parameters used in this initial analysis. This makes it easy to systemize the identification of customers with a disadvantageous heat load pattern for both the customers and the district heating companies. 17

Thirdly, the developed method can probably be enhanced by redefining the customer categories by their indoor activities. The best example is Health and Social Services buildings that should be split into groups depending on the activity and the duration of activity in the buildings. 5. Acknowledgements This analysis was performed by the financial support from Fjärrsyn, the Swedish district heating research programme, and Öresundskraft, which also provided the time series for the analyses. References [1] Wissner M, The Smart Grid-A saucerful of secrets? Applied Energy 88 (2011) 2509-2518. [2] Aronsson S, Fjärrvärmekunders värme- och effektbehov, (Heat and power demands for district heating customers), Doctoral thesis, Department of Building Services Engineering, Chalmers University of Technology, Gothenburg, 1996. [3] Böhm B, Daniig P O, Monitoring the energy consumption in a district heated apartment building in Copenhagen, with specific interest in the thermodynamic performance. Energy and Buildings 36 (2004) 229-236. [4] Di Perna C et al, Influence of the internal inertia of the building envelope on summertime comfort in buildings with high internal heat loads. Energy and buildings 43 (2011) 200-206. [5] Kiluk S, Algorithmic acquisition of diagnostic patterns in district heating billing system. Applied Energy 91(2012) 146-155. [6] Gustafsson J et al, Improved district heating substation efficiency with new control strategy. Applied Energy 87 (2010) 1996-2000. [7] Lauenburg P, Improved supply of district heating to hydro space heating systems. Doctoral thesis, Department of Energy Sciences, Faculty of Engineering, Lund University, Lund 2009. [8] Gustafsson J et al, Experimental evaluation of radiator control on primary supply temperature for district heating substations. Applied Energy 88 (2011) 4945-4951. [9] Winberg A, Werner S, Avkylning av fjärrvärmevatten i befintliga abonnentcentraler, (Cooling of the District Heating Fluid in Existing Consumer Substations), Stiftelsen för värmeteknisk forskning, Stockholm, 1987. [10] Yliniemi K, Fault detection in district heating substations. Licentiate thesis, Department of Science and Electrical Engineering, Luleå University of Technology, Luleå 2005 [11] Statistics Sweden, El-, gas- och fjärrvärmeförsörjningen 2009 (Electricity, gas and district heating supply 2009), Statistiska meddelanden EN11 SM1101. [12] Gadd H, Werner S, Daily variation in Swedish district heating systems. The 12th International Symposium on District Heating and Cooling, Tallinn 2010. [13] Frederiksen S, Werner S, Fjärrvärme - Teori, teknik och funktion (District heating - Theory, Technique and Function), Studentlitteratur, Lund 1993. 18