THE APPLICATION OF AN INDUSTRIAL DEMAND-SIDE MANAGEMENT EXPERT SYSTEM FOR THE ANALYSIS OF ELECTROHEAT TECHNOLOGY RETROFITS

Transcription

1 THE APPLICATION OF AN INDUSTRIAL DEMAND-SIDE MANAGEMENT EXPERT SYSTEM FOR THE ANALYSIS OF ELECTROHEAT TECHNOLOGY RETROFITS R A Harper (1), J E Gilmour (1), T N Oliver (1), M Booth (2) (1) Aston University, United Kingdom (2) Energy Services UK Ltd, United Kingdom SUMMARY The paper details the development of a Decision Support System for the process of electro-technology retrofits, within industrial sub-sector Demand-Side Management (DSM) programmes. The completion of an industrial drying energy audit concluded that the manufacturing process accounts for approximately 18% of total industrial energy consumption within the United Kingdom (UK). The energy consumption of drying operations was also found to be increasing in relation to total industrial energy use. The energy intensive nature of the process is highly suited to subsector DSM programmes, particularly in areas of high dryer concentrations, for example, the UK ceramic manufacturing sector. The nature of drying technology presents opportunities for targeting energy inefficient processes and for proposing solutions through load management programmes, incorporating various levels of technical procurement. A knowledge-based simulation has been developed that allows for the direct comparison of a range of drying technologies and for each, the associated energy expenditure, atmospheric emissions and manufacturing implementation. It is intended that this will form a single module within a significantly larger Integrated Energy Programme (IEP), directly targeted towards industrial process opportunities. the derivation of approximate values for total energy consumption, cost and carbon dioxide emission levels. The system provides simulation assessment of alternative energy efficient technologies in direct comparison to existing equipment, thus allowing analysis and forecasting of potential impacts on the energy distribution network. A case study from the UK ceramics manufacturing sector suggests a potential 15% to 40% energy saving through the implementation of new drying technology. An average value of 36% of total site electrical load can be apportioned to drying operations. The figure is significant in terms of the application of traditional DSM measures, particularly when the majority of electrical load is due to motive devices, predominantly burner, re-circulation and exhaust fans. The fuzzy logic Decision Support System has proved applicable for the approximate modelling of dryer behaviour. The paper demonstrates the potential for using an Artificial Intelligence technique for marketled energy efficiency programmes within industrial sub-sectors, particularly in situations where complex process modelling is necessary. On a production basis it is difficult to model Specific Energy Consumption (SEC) values because dryer output does not often produce components for which direct sales or production figures exist. In addition, drying is not necessarily the primary energy consumer in most sectors. Calculation of energy consumption at the process level utilises both time and capital through energy audit implementation. The data generated is highly specific to the process audited. Hence there is a need for a more generic modelling strategy based upon drying times, drying loads, and approximate dryer characteristics. The drying model must also account for indirect energy use, including the technologies of humidification systems, and motive power. The authors have identified a potential application for the use of a Fuzzy Inference System (FIS) as a tool to predict drying rates and subsequent drying times. The combination of such data together with material properties and technical drying characteristics allow for

2 L'APPLICATION D'UN SYSTÈME EXPERT DE GESTION INDUSTRIELLE DES BESOINS POUR L'ANALYSE DES MISES À NIVEAU A POSTERIORI UTILISANT LA TECHNOLOGIE DE L'ÉLECTROCHALEUR R A Harper (1), J E Gilmour (1), T N Oliver (1), M Booth (2) (1) Université d' Aston, Royaume-Uni (2) Energy Services UK Ltd, Royaume-Uni RESUME Ce document détaille le développement d'un Système d'aide à la Décision pour le procédé des mises à niveau a posteriori utilisant l'électro-technologie, au sein des programmes de Gestion Industrielle des Besoins de sous-secteur (GIB). La réalisation d'un audit d'énergie de séchage industriel a conclu que les procédés de fabrication utilisent environ 18% de la consommation d'énergie industrielle totale au Royaume-Uni (R-U). Il a aussi été découvert que la consommation d'énergie des opérations de séchage augmente en fonction de l'utilisation d'énergie industrielle totale. La nature de procédé à grande consommation d'énergie est très adaptée aux programmes de GIB des sous-secteurs, spécialement dans les secteurs comportant des concentrations de sécheurs élevées, comme, par exemple, le secteur de la fabrication de la céramique au Royaume-Uni. La nature de la technologie du séchage présente des occasions de cibler les procédés à rendement énergétique médiocre et de proposer des solutions par l'intermédiaire de programmes de gestion des charges, incorporant divers niveaux d'approvisionnement technique. Une simulation basée sur la connaissance a été développée, permettant la comparaison directe d'une gamme de technologies de séchage, et, pour chacune, la comparaison directe des frais d'énergie, des émissions atmosphériques et de l'implémentation dans la fabrication. Il est projeté que ceci forme un simple module au sein d'un Programme d'energie Intégré (PEI), directement ciblé vers les perspectives des procédés industriels. Du point de vue de la production, il est difficile de modeler des valeurs de Consommation d'energie Spécifique (CES), parce que la productivité des sécheurs ne produit pas souvent des composants pour lesquels des ventes directes ou des chiffres de production existent. De plus, le séchage n'est pas nécessairement le consommateur d'énergie primaire dans la plupart des secteurs. Le calcul de la consommation d'énergie au niveau des procédés utilise du temps et des capitaux, dû à une mise en œuvre comportant des audit d'énergie. L'information générée est hautement spécifique au procédé audité. Il existe donc le besoin d'une stratégie de modélisation plus générique basée sur les temps de séchage, sur les charges de séchage, et sut les caractéristiques approximatives des sécheurs. Il est aussi nécessaire que les modèles de séchage prennent en compte l'utilisation indirecte de l'énergie, comportant les technologies des systèmes d'humidification et la puissance motrice. Les auteurs ont identifié une application potentielle pour l'utilisation d'un Système à Inférence Floue (SIF) comme un outil de prédiction des vitesses de séchage et des temps de séchage subséquents. La combinaison d'une telle information avec les propriétés de matière et avec les caractéristiques techniques de séchage permet de calculer les valeurs approximatives de la consommation d'énergie totale, des frais et des niveaux d'émission de gaz carbonique. Le système procure l'évaluation par simulation des technologies à rendement énergétique efficace en comparaison directe avec le matériel existant, permettant ainsi l'analyse et la prévision des impacts potentiels sur le réseau de distribution d'énergie. Une étude de cas du secteur de fabrication de la céramique au Royaume-Uni indique une économie d'énergie potentielle de 15 % à 40 % grâce à la mise en œuvre de la nouvelle technologie de séchage. Il est possible d'assigner une valeur moyenne de 36% de la charge électrique totale d'un site aux opérations de séchage. Ce chiffre est important dans des termes d'application des mesures de GIB, spécialement lorsque la plupart des charges électriques est due aux dispositifs moteurs, principalement les ventilateurs de four, les ventilateurs de re-circulation et les ventilateurs d'échappement. Le Système d'aide à la Décision à Inférence Floue a été prouvé applicable pour la modélisation approximative du comportement des sécheurs. Le document explique le potentiel de l'utilisation d'une technique d'intelligence artificielle pour les programmes de rendement énergétique menés par le marché dans les sous-secteurs industriels, spécialement pour les situations dans lesquelles des modélisations de procédé complexes sont nécessaires.

3 THE APPLICATION OF AN INDUSTRIAL DEMAND-SIDE MANAGEMENT EXPERT SYSTEM FOR THE ANALYSIS OF ELECTROHEAT TECHNOLOGY RETROFITS R A Harper (1), J E Gilmour (1), T N Oliver (1), M Booth (2) (1) Aston University, United Kingdom (2) Energy Services UK Ltd, United Kingdom INTRODUCTION Energy efficiency at a process level is gaining increasing importance with regard to industrial technology procurement. Legislative developments in the United Kingdom (UK) have proposed the implementation of the Climate Change Levy (CCL) from April The CCL will establish the backbone of the UK Government s commitment towards the reduction in carbon dioxide emissions as agreed under the 1997 Kyoto Protocol. The implications of the CCL are as yet unknown, however Department of the Environment, Transport and the Regions (DETR) consultation documents have proposed a unit cost increase on electricity, gas and oil, Department of the Environment (2). Energy intensive sectors have acknowledged the need to reduce energy consumption through the implementation of long term efficiency strategies. There is evidence to suggest that Energy Service Company s (ESCO) may operate within the legislative restrictions of a liberalised energy market in order to provide expert knowledge regarding the efficient consumption of energy. The CCL and the potential of future ethical drivers provide opportunities for ESCO involvement and interaction throughout industrial energy efficiency strategies. Previous studies have demonstrated the high levels of energy consumption associated with industrial drying processes, Jay (9). The main reasons for this are: is energy intensive by nature, processes are prevalent within industrial environments, technologies display inherent inefficiencies; and, There are often historical reasons for the choice of process technology. The nature of drying technology presents opportunities for targeting energy inefficient processes and for proposing solutions through load management programmes, incorporating various levels of technical procurement. A knowledge based simulation has been developed that allows for the direct comparison of a range of drying technologies and for each, the associated energy costs, atmospheric emissions and manufacturing implementation. It is intended that this will form a single module within a significantly larger Integrated Energy Programme (IEP), directly targeted towards process improvement opportunities. INDUSTRIAL DSM IN COMPETITIVE UTILITY MARKETS The concept of Demand-Side Management (DSM) provides energy utility companies with a strategic focus on energy utilisation through a Least Cost Planning (LCP) initiative. DSM originated in the United States (US) during the 1970 s, and has been a successful option available under LCP. By 1992 the US energy industry had invested in excess of $2,550 million in DSM programmes, Steer (14). The structure of the UK energy markets, particularly the Electricity Supply Industry (ESI) presents a series of barriers to utility DSM investment. Clarification and expansion regarding the precise nature of the reasons behind the inappropriate transition of utility DSM from US to UK markets can be found in (14) and Eyre (4). In the main, the economic business drivers associated with utilisation projects have altered as a result of environmental, legislative and market structure changes. A potential solution enabling the application of DSM programmes within the UK concentrates investment within a particular industrial sub-sector, Steer (15). This concept of a Market-Led DSM strategy is a significant step away from the traditional utility DSM programmes, and is allied towards energy services, Gilmour (5). Figure 1 indicates the prioritisation of activities for industrial DSM measures as determined by the Swiss energy industry s working group on DSM. The diagram describes the drying process as a first tier business opportunity. The importance of such a relationship becomes apparent when DSM strategic planning is analysed with regard to energy market liberalisation. A competitive energy supply market has been established in the UK through a series of incremental steps, based upon maximum load consumption. A primary stage within this liberalisation has been the separation of the distribution and supply divisions of the former energy companies. This has resulted in the unsuitability of utility DSM within the UK ESI due to

4 legislative restrictions of end-use load consumption data. Promotion of supply competition has had the effect of increasing the number of supply companies operating across the franchise region of a single distribution network, thereby complicating the central planning strategy. The inherent nature of a competitive market has led to a reduction in the unit cost of energy. Although this produces a negative effect on industrial efficiency investment (4), potential business opportunities arise for energy efficiency programmes implemented and managed through ESCO s. Process heat Heating Cooling Motors Lighting Other Commerce Industry Services 1 st priority 2 nd priority Other Figure 1-Market-led Industrial DSM Options (Source: VSE Working Group on DSM, DSM Summary and Recommendations, 1998). The proposed solution is through the implementation of industrial energy efficiency projects aimed at directly benefiting the consumer. There is now a reduced emphasis on load reduction programmes for energy distribution companies, and a transition towards value-added supply contracts through ESCO involvement. Although the conceptual form of DSM remains an important factor for distribution companies, the business drivers no longer exist through which to operate utility load management programmes. UK INDUSTRIAL DRYING ENERGY CONSUMPTION The analysis of energy consumption for drying is complicated. The authors suggest that generalised modelling is rare for a number of reasons, including: The different energy sources available (steam, gas, oil, electroheat), The operation of different dryer types (direct, indirect, spray, chamber etc.); and, The number of dryer configurations (dryer size, burner type, motive devices). On a production basis it is difficult to model Specific Energy Consumption (SEC) values because dryer output does not often produce components for which direct sales or production figures exist, Holmes (8). In addition, drying is not necessarily the primary energy consumer in most sectors. Therefore, any requirement to model energy consumption is of a low priority. Hence, drying technology investment is not conducted using optimised data levels. Two recent publications have attempted to estimate total energy use for industrial drying within the UK, Gilmour (6), Jay and Oliver (10). A further five previous studies have been undertaken between the years 1976 and 1989, Hodgett (7), Witt (18), ETSU (3), Baker and Reay (1), Wilmshurst (17). Audit analysis suggests that energy use for industrial drying can be estimated between 5.5% and 18.2% of total UK industrial energy consumption. More importantly Jay acknowledged the trend that energy for drying operations was found to have an increasing relationship against total industrial energy use (figure 2). Energy for (PJ) Year Figure 2-Variation in energy use for UK industrial drying between 1978 and 1990, (9). Although highlighting trends in energy use, previous estimations of energy consumption for drying on a national scale have used differing calculation and assumption methodologies. Subsequently, there is little commonality between individual surveys. Moreover, although industry-wide surveys provide a good indication of energy use across the UK, they provide insufficient detail for use at the process level. Calculation of energy consumption at the process level utilises both time and capital through energy audit implementation. The data generated is highly specific to the process audited. Hence there is a need for a more generic modelling strategy based upon drying times, drying loads, and approximate dryer characteristics. The drying model must also account for indirect energy use, including the technologies of humidification systems, and motive power. A search of literature highlights the recent application of Artificial Intelligence (AI) techniques, specifically Artificial Neural Networks (ANN) for the modelling of

5 drying behaviour, Jay and Oliver (11). The neural network systems investigated require no formal mathematical modelling or irreversible thermodynamic equations. Models have been derived and instructed purely from empirical data. The benefits and limitations of using neural networks for the modelling of drying processes are described in table 1. Benefits General learning capability Avoidance of firstprinciples models Reduced development costs Reduced development time Potential for adaptive solutions Disadvantages Continuing risk of overfitting data, thus highlighting unwanted trends within data sets A substantial degree of specialised knowledge is required for neural network engineering and programming. ANN systems are not particularly intuitive from the perspective of the end-user Good performance is highly dependent upon the interconnection model used and how well it models the problem, Sutton (16) Table 1-Summary of ANN limitations for the modelling of drying processes. Due to the ANN requirement for specialised programming knowledge the resultant modelling capabilities are often dependant on the network structure employed. Additionally, when used within an IEP there is an inability to view the actual modelling process. Consequently this may lead to a lack of customer confidence in the ESCO decision support system. A solution is proffered through the use of fuzzy systems. FUZZY LOGIC DECISION SUPPORT SYSTEM A fuzzy model is essentially a black box in which input data is mapped to output data. Fuzzy inference is simply the process of mapping using fuzzy logic, or classifying objects separated by unclear boundaries, in which membership is not black and white but a matter of degree. Fuzzy systems have gained increasing popularity in engineering over the past few decades, finding a large variety of applications in control theory, pattern recognition, power systems and expert prediction systems, Roger Jang and Gulley (13). The benefits of fuzzy systems include: Rapid calculation times. Fast and inexpensive construction. Easy to understand. Intuitively simple to operate. Flexibility in up dating. Tolerant of imprecise data. The modelling of non-linear systems. In addition to the above advantages, fuzzy models can be combined with neural networks to create Adaptive Neuro Fuzzy Inference Systems (ANFISs). The authors have found limited published work in relation to the modelling of energy use and energy efficiency for drying processes. Furthermore, inclusive of the work that exists, no information has been found which specifically demonstrates the modelling of drying processes for the application of energy efficiency programmes. The authors have identified a potential application for the use of a Fuzzy Inference System (FIS) as a tool to predict drying rates and subsequent drying times. The combination of such data together with product thermophysical properties and technical drying characteristics allowed the derivation of approximate values for total energy consumption, cost and carbon dioxide emission levels. It is the author s believe that the work presented in the paper is the first to investigate the modelling of drying rates and drying times using a fuzzy system. Three fuzzy models have been derived, based upon the technologies of conventional (convective) air drying, dehumidification heat pump drying and airless drying. Models were instructed using empirical data from the drying of electro-porcelain insulator material. To provide realistic and practical models for the prediction of drying rates, it is necessary to base the models on input variables, which are both measurable and controllable in an industrial situation. Contact with the UK ceramics industry indicates that air temperature and relative humidity (rh) are the fundamental control variables when drying ceramic components. Air and heat pump drying models therefore use a model scheme based on air temperature and relative humidity, whereas airless drying uses a single input directly related to steam temperature. Model outputs are presented as an overall drying rate, defined as the rate of mass loss per unit area of material (g/m 2 /s). Modelling data was obtained from the controlled drying of samples in laboratory-scale dryers. Air drying and heat pump drying was performed using a range of temperatures from 30 C to 90 C, and 30 C to 70 C respectively. Relative humidity control varied from a no control scenario to 40%. Component samples were airless dried using steam temperatures between 120 C and 150 C. Moisture loss throughout the drying cycles was measured using electronic balances linked to a PC data acquisition system. The termination of drying was determined by measurement of product weight and core temperature. Input and output data was mapped using a purpose written

6 Matlab programme, and all modelling was performed using the Matlab Fuzzy Logic Toolbox. Figure 3 presents the ANFIS output surface space for the heat pump drying of electro-porcelain insulating material. The surface space is derived directly from the fuzzy rules in the drying model. Using this model, any combination of drying temperature (x-axis) and relative humidity (y-axis) within the experimental range will derive an estimated drying rate as an output (z-axis). Table 2 presents model output data which provides evidence that airless drying exhibits a significantly higher drying rate compared to both heat pump and convective air drying. Overall, airless drying would reduce process time by 52 hours. The use of heat pump drying would extend drying time by 3.6 hours compared to current drying practice, although energy costs would be reduced by approximately 45%. As expected, thermal loads are identical across all the models. discipline load (kg) rate (g/m 2 /s) time (hrs) Gas to Elec. Energy (%) Convective :43 Airless :15 Heat pump :100 Table 2-Fuzzy models drying rate comparison. Figure 3-Output surfaces diagram (Heat Pump Dehumidification). The fuzzy models were incorporated within the Matlab Simulink Toolbox to determine a set of simplified IEP models for the heat pump dehumidification, conventional air and airless drying procedures. When combined within a single module, a specific input produced three separate outputs relating to the alternative drying technologies. A direct comparison is then possible when analysing total energy consumption, cost and carbon dioxide emissions. CASESTUDY: DRYING ELECTRO-PORCELAIN The situation exits whereby a manufacturer of ceramic insulators runs 2 convective air dryers simultaneously for 42 weeks of the year. The individual dryers are of modern design (operating at an assumed overall efficiency of 50%). The dryers are capable of processing 400 components having dimensions of 152mm diameter and 380mm length. The components have a wet weight of 10kg and moisture content of 30 per cent dry weight base (dwb). Assuming the current drying schedule proceeds at an average of 65 C and 40% rh, the question can be asked what are the potential benefits from using airless dryers for the drying time? Alternatively, what benefits could be realised from introducing heat pump dehumidifiers into the drying schedule? The operating times for dryers are dependant upon production scheduling and manufacturing policy. The shorter processing times associated with batch dryers are advantageous when considering flexibility during production. Figure 4 presents a comparison between the electrical energy profiles for the convective air, heat pump and airless drying models. Electricity Demand (kwh) :00 02:30 04:00 05:30 07:00 08:30 10:00 11:30 Time (hrs) 13:00 14:30 16:00 17:30 19:00 20:30 22:00 23:30 Convective Air Heat Pump Airless Figure 4-Electrical demand comparison for dryer types. The profiles represent electrical loads for the drying of a similar mass of porcelain. It is apparent that convective air drying consumes significantly more electrical energy than heat pump dehumidification or airless drying. Airless drying demonstrates a lower dependence on electricity than conventional drying. It can be suggested that the higher thermal capacity of superheated steam results in the requirement for reduced fan size and power, Maloney (12). On a cost basis, all drying disciplines demonstrate significant use of electrical energy. SEC values and energy profiles provide useful tools into the effectiveness of energy use at the process level.

7 However, for Energy Services and DSM it is imperative that the relative efficiencies of the different drying technologies are considered. Analysis of dryer efficiency allows an engineer or energy manager to assess the main source of energy inefficiency, thereby highlighting issues of energy consumption. Problems can then be targeted within load management programmes through energy service initiatives. CONCLUSIONS The fuzzy logic Decision Support System has proved applicable for the approximate modelling of dryer behaviour. This paper has demonstrated the potential for using an Artificial Intelligence technique for market-led energy efficiency programmes within industrial sub-sectors, particularly in situations where complex process modelling is necessary. The analysis of drying technology for the reduction of emission levels and energy consumption has both ethical and business drivers. Airless and heat pump drying technology provide evidence of the possible energy consumption benefits gained through electrotechnology retrofits. This has beneficial implications when considering the predicted effect of the Climate Change Levy. The authors recognise the requirement for further research in order to prove the system for a range of electro-technologies used for energy efficiency programmes throughout additional industrial sub-sectors. REFERENCES 1. Baker, C. G. J. and Reay, D., 1982, September, Energy Usage for in Selected UK Industrial Sectors. Proc. 3rd Int. Symposium, IDS 82, Birmingham, UK 2. Department of the Environment, T. a. t. R., 2000, March, Energy Efficiency Measures under the Climate Change Levy Package, x.htm. 12/3/00 3. Energy Technology Support Unit (ETSU), 1985,, Evaporation and Distillation: The Potential for Improving Energy Efficiency in Twelve Industrial Sectors. ETSU Market Study,3, Energy Publications/Energy Efficiency Office, UK 4. Eyre, N., 1998, A Golden Age or a False Dawn? Energy Efficiency in UK Competitive Energy Markets, Energy Policy, 26,12, Gilmour, J. E., Steer, G. C., Oliver, T. N. and Booth, M, 1998, September, Energy Utilisation within the UK Sanitaryware Industry and the Potential for Energy Services. Universities Power Engineering Conference, UPEC 98, Edinburgh, UK 6. Gilmour, J., Oliver, T. N. and Booth, M, 1998, August, Energy Use for Processes: The Potential Benefits of Airless, Proc. 11th Int. Symposium, IDS 98, Halkidiki, Greece, Hodgett, D. L., 1976, July/August, Efficient Using Heat Pumps, The Chemical Engineer. 8. Holmes, J. G., Hedman, B. A. and Salama, S. A., 1988, July, Overview of Industrial Needs and Competing Technologies, Plant & Operations Progress, 7,3, Jay, S., 1996, Processes in the United Kingdom, diss. Birmingham, UK, Aston University 10. Jay, S. and Oliver, T. N., 1994, August, Energy Consumption for Industrial Processes in the United Kingdom, Proc. 9th Int. Symposium, IDS 94, Gold Coast, Australia 11. Jay, S. and Oliver, T. N., 1996, 30th July-2nd August, Modeling and Control of Processes Using Neural Networks, Proc. 10th Int. Symposium, IDS 96, Krakow, Poland 12. Maloney, N. J., 1995, March, Mechanical Vapour Recompression and Superheated Steam, Novel Techniques for Process Advantage. IChemE Solids Group Seminar, EA Technology, Capenhurst, UK 13. Roger Jang, J. S. and Gulley, N., 1996, Fuzzy Logic Toolbox: Reference Manual, The Mathworks Inc. 14. Steer, G., 1999, An Energy Services Approach to Industrial Sector Demand Side Management, diss. Aston University, Birmingham, UK 15. Steer, G. C., Oliver, T. N. and Booth, M., 1997, September, The Potential for Demand-Side Management within a Specific Industrial Sub-Sector, Universities Power Engineering Conference, 61-64, UPEC 97, UMIST, Manchester, UK 16. Sutton, J. C., 1992, Manufacturing Applications of Neural Networks for the 90s, Proc. Manu. Int., , MI 92, Dallas, USA 17. Wilmshurst, A., 1989, Industrial with Special Reference to Electric Infra-red, diss. Cambridge, UK 18. Witt, J. A., 1977, Market Information, Electricity Council Marketing Department, London, UK