Energy viability of photovoltaic systems

Transcription

1 Energy Policy 28 (2000) 999}1010 Energy viability of photovoltaic systems E.A. Alsema*, E. Nieuwlaar Department of Science, Technology and Society, Utrecht University, Padualaan 14, NL-3584 CH Utrecht, The Netherlands Received 24 May 2000 Abstract The energy balance of photovoltaic (PV) energy systems is analysed in order to evaluate the energy pay-back time and the emissions of grid-connected PV systems. After a short introduction on energy analysis methodology we discuss the energy requirements for production of solar cell modules based on crystalline silicon and on thin-"lm technology, as well as for the manufacturing of other system components. Assuming a medium}high irradiation of 1700 kwh/m yr the energy pay-back time was found to be 2.5}3 yr for present-day roof-top installations and almost 4 yr for multi-megawatt, ground-mounted systems. Prospects for improvement of the energy balance of PV systems are discussed and it is found that for future PV technology (in 2020) the energy pay-back time may be less than 1.5 yr for roof-top systems and less than 2 yr for ground-mounted systems (under the same irradiation). The speci"c emission of the roof-top systems was calculated as 50}60 g/kwh now and possibly around 20 g/kwh in the future. This leads to the conclusion that emissions of present PV systems are considerably lower than emissions from fossil-fuel power plants, but somewhat higher than for wind and biomass energy. No signi"cant contribution to mitigation should be expected from PV technology in the year In the longer term, however, grid-connected PV systems do have a signi"cant potential for mitigation Elsevier Science Ltd. All rights reserved. Keywords: Solar power; Energy pay-back time; mitigation 1. Introduction Photovoltaic energy conversion is widely considered as one of the more promising renewable energy technologies which has the potential to contribute signi"cantly to a sustainable energy supply and which may help to mitigate greenhouse gas emissions. For this reason there is a growing support from governments for photovoltaic R&D programmes and market introduction schemes. In order to ful"l these promises photovoltaic (PV) technology has to meet two requirements: (1) PV energy generation should have an acceptable cost/performance ratio and (2) the net energy yield for PV systems should be (much) larger than zero. With a positive energy yield we mean that the energy output during the lifetime of the PV system must be larger than the energy inputs during the system's life cycle, i.e. for manufacturing of the components and for the installation, maintenance and decommissioning of the PV sys- * Corresponding author. Tel.: # ; fax: # address: e.a.alsema@chem.uu.nl (E.A. Alsema). tem. Of course, the data on net energy yields have their implications for the evaluation of the cost/performance ratios, as performance should be based on net yields instead of gross yields. In practice this is not always done which may lead to somewhat over-optimistic evaluations of the economic viability of PV systems. Also projections of the mitigation potential of PV technology should be based on expected net energy yields. The net energy yield may also have some predictive value for the feasibility of an acceptable cost/performance ratio, which we mentioned above as the "rst requirement. Namely, if PV would have a very low net energy yield, there is a good chance that it will be di$cult to achieve a fair cost/performance ratio. In our view every new energy technology which is promoted as being `renewablea or `sustainablea should be subjected to an analysis of its energy balance in order to asses its `energy viabilitya and to calculate the net energy yield. Of great importance is that such an energy analysis is not only based on data for present-generation systems but also considers expected improvements in production and energy system technology. Since energy consumption generally has signi"cant environmental implications, the energy analysis may be considered as /00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S ( 0 0 )

2 1000 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999}1010 a "rst step towards a more comprehensive environmental life-cycle assessment. Furthermore, energy analysis results provide a good indication of the mitigation potential of the considered energy technology. Our objective in this paper is to present estimates of the energy requirements for manufacturing of PV systems and to evaluate the energy balance for a few representative examples of PV system applications. We will also investigate the e!ects of future enhancements in PV production technology and PV system technology in order to assess the long-term prospects of PV technology as a candidate for a sustainable energy supply and for mitigation. We will restrict our evaluations to PV systems that are connected to an electricity grid. Stand-alone systems, like the so-called `Solar Home Systemsa, are left out of consideration here, because these systems have a much smaller potential in terms of long-term, global energy supply and because energy analysis results on standalone systems are hardly available at present. First we will give a short discussion on energy analysis methodology (Section 2), and then we will apply these methods to the components of PV systems (Section 3). In Section 4 we present the energy balance for grid-connected PV systems, followed by an outlook for future PV systems in Section 5. In Section 6 the potential for mitigation using PV systems will be assessed. Finally policy implications (Section 7) and conclusions (Section 8) are given. 2. Energy analysis methodology In an energy analysis a comprehensive account is given of the energy inputs and outputs involved in products or services. The overall energy performance of such products and services is determined by accounting all energy #ows in the life cycle (from resource extraction through manufacturing, product use until end-of-life decommissioning). In the case of solar cells, the gross energy requirement E is determined by adding together the energy input during resource winning, production, installation, operation and decommissioning of the solar cell In a life-cycle assessment (LCA) an inventory is made of all material and energy requirements, as well as emissions to the environment, after which the inventory results may be evaluated regarding their environmental impacts. Full LCAs are di$cult to perform for technologies which are still in an R&D phase or in early stages of implementation, because reliable data on emissions, etc. are lacking. Energy analyses are somewhat simpler and more reliable because energy consumption data are more readily available. An energy analysis of Solar Home Systems is currently underway at our department, as part of a full life-cycle assessment study. The solar energy input is normally left out, since it is not considered as a commercial energy carrier (i.e. the solar energy is a `freea good). In fact, the calculation of energy payback times and energy yield factors depends on not counting solar energy input. panels and the other system components. This gross energy requirement is then compared with the energy output E. Two indicators for the overall energy performance of solar systems can be derived from this information. First, an energy pay-back time (EPBT) can be calculated by dividing the gross energy requirement (E ) by the annual energy output (E ). The energy payback time shows how long it takes before energy investments are compensated by energy yields. Since only part of the total energy yield is re#ected in the energy payback time, it is useful to present also the net energy balance by calculating an energy yield factor by dividing the lifetime energy output E by the gross energy requirement. The energy yield factor shows how much energy is obtained per unit `investeda energy. However, because energy yield factors do not provide much more insight than energy pay-back times we will mainly employ the latter indicator. (In fact the energy yield can be obtained very easily by diving the lifetime by the EPBT.) A number of factors must be considered when calculating and presenting energy payback times and energy yield factors: Which processes in the life cycle are evaluated regarding their energy inputs and outputs? This is dealt with by specifying the system boundary which delimits the processes analysed from the processes that are not accounted for. The reasons for not including some processes in the system boundary are either that they can be considered as negligible or that reliable data are lacking. Some processes have negligible energy requirements. Often the energy requirement for manufacturing capital goods is considered as negligible. For PV power systems however the energy requirement for capital goods (e.g., machines used to produce modules) is signi"cant and will therefore be accounted for in our analyses. The procedure and data used to estimate the energy requirement for capital equipment are explained in Alsema (1998). It involves the use of energy input}output analysis of economic sectors leading to cumulative energy intensities for the manufacture of capital equipment expressed in energy per currency unit (e.g., MJ/dollar). Combining this result with the investment cost for a PV plant results in an estimate for the energy requirement for capital goods. Other processes are not re#ected in the calculations because of lack of data. In our case, end-of-life energy #ows, e.g., for module recycling, are not counted because of uncertainties and lack of data. How will secondary energy carriers (notably electricity) be treated in the analysis? All energy use in the life cycle of PV systems must be calculated back to the extraction of primary energy

3 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999} carriers. The results of the energy analysis are in particular a!ected by the e$ciency of electricity production. The conversion e$ciency used here is 0.35, which is an often used average value. This factor is also used to calculate the primary energy equivalent of the electricity produced by the PV power system. Which common property will be used in the presentation of energy requirements for PV modules? We have chosen to analyse and present the energy requirements in megajoules per square metre module area. In general, the energy requirements for module production are area dependent and not power dependent. Relating the energy requirements to module output (Wp or kwh) introduces extra parameters, namely module e$ciency, PV system performance and irradiation, parameters which we prefer to treat seperately. The advantage of this approach is that energy requirement data for modules with di!ering e$ciencies can be analysed and compared much better. Only after we have established the area-related energy data, we will factor in the power output per square metre module area in order to obtain the energy requirements per Wp. As a third step system performance and irradation assumptions can be taken into account to evaluate the energy balance on a system level. 3. Energy requirements of PV systems 3.1. General Over the past decade a number of detailed studies on the energy requirements for PV modules or systems have been published (see for example, Hagedorn and Hellriegel, 1992; Phylipsen and Alsema, 1995; Alsema, 1996; Nijs et al., 1997; Keoleian and Lewis, 1997; Frankl et al., 1998; Kato et al., 1998). We have reviewed and compared these and other studies and tried to establish on which data there is more or less consensus and how observed di!erences may be explained. Based on this review of available data we have established a `best estimatea of the energy requirement of crystalline silicon modules, thin "lm modules and other system components (Alsema et al., 1998; Alsema, 1998). Here we will present only the main results of these analyses, for a more detailed breakdown of the solar cell production process the reader is referred to the aforementioned publications. Also the reasons for the relatively high uncertainty (&40%) in the estimates can be found there. Module power ratings are de"ned at an irradiation of 1000 W/m and expressed in Watt-peak (Wp) Crystalline silicon modules First we will discuss crystalline silicon module technology, which is presently the dominant technology in the market. The energy conversion e$ciency is around 13% for multicrystalline silicon and around 14% for monocrystalline silicon modules. Before we can start with our energy analysis we will shortly outline the production process for crystalline silicon modules. Silicon is produced from silica (SiO2) which is mined as quartz sand. Reduction of silica in large arc furnaces yields metallurgical-grade silicon, which has to be puri"ed further to electronic-grade silicon before it is suitable for manufacturing of electronic components such as integrated circuits or solar cells. Moreover, both the electronics industry and the PV industry use silicon in a (mono)crystalline form. Therefore, the silicon is molten and subsequently crystallized under carefully controlled conditions so that large blocks or ingots of crystalline material are obtained. After removal of the outer edges of the ingot (material of insu$cient quality), the ingots are sawn into smaller blocks and then into thin slabs or wafers. Typical wafers are 0.3 mm thick and 100 cm in area. Wafers of either monocrystalline or multicrystalline silicon are the starting point for the actual solar cell production, which involves several steps that we will not explain here. When the solar cells are "nished they are laid out into a matrix of for example 9 4 cells, interconnected and encapsulated between a glass front plate, a lamination foil and a back cover foil. In Table 1 we show the energy requirements for a typical module based on multicrystalline silicon. We can see that the silicon winning and puri"cation process is the major energy consumer and responsible for about half of the module's energy requirement. The reason for this is that the puri"cation is performed by a highly energyintensive process and the purity criteria are laid down by the electronics industry, which consumes about 90% of the `electronic-gradea silicon (Bruton et al., 1996). Although solar cell production would also be possible with more relaxed purity standards, dedicated puri"cation of `solar-gradea silicon has up to now not been commercially feasible because of the relatively small demand for this type of silicon material. The production of the silicon wafer from the electronic-grade silicon also has a signi"cant contribution to the energy consumption, mainly because 60% of the material is lost during the formation of the multicrystalline silicon ingots and the sawing of these ingots into Multicrystalline material is not of a perfect crystalline form. It can be produced more easily and thus at lower costs. However, the conversion e$ciency of modules employing this material is generally 1}2 percent-point lower than with monocrystalline silicon.

4 1002 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999}1010 Table 1 Breakdown of the energy requirements for a typical multicrystalline silicon PV module using present-day production technology (in MJ of primary energy per m module area) Process Silicon winning and puri"cation 2200 Silicon wafer production 1000 Cell/module processing 300 Module encapsulation materials 200 Overhead oper. and equipment manuf. 500 Total module (w/o frame) 4200 Module frame (aluminium) 400 Total module (framed) 4600 Energy requirements (MJ /m module) wafers (Nijs et al., 1997). Altogether the production of the silicon wafers requires about 3200 MJ/m module area, which is about 60% of the total energy requirement for the module. For monocrystalline silicon wafers the more elaborate crystallization process increases the energy requirements with an additional 1500 MJ/m. The energy requirements for cell and module processing are more modest (300 MJ/m ) just as the energy requirements for the front cover glass, the lamination foil and the back cover foil (200 MJ/m ). A typical aluminium frame around the module may add an extra 400 MJ/m, but frame materials and dimensions can vary considerably. In fact, frameless modules are also available on the market. Also note the relatively high energy use for overhead operations (climatization of production facility, lighting, compressed air, etc.) and for the manufacturing of the production equipment itself (together 500 MJ/m ). The latter is due to relatively high capital costs of solar cell manufacturing plants (Bruton et al., 1996), costs which also formed the basis for our energy input estimate (see Section 2). Based on a conversion e$ciency of 13%, we have a power output of 130 Watt-peak per m module, so the energy requirements of present-day mc-si modules can be evaluated as about 32 MJ/Wp without frame and 35 MJ/m with an aluminium frame Thin-xlm modules Regarding thin-"lm modules a number of contending cell types and production technologies exist and as yet it is not clear which one of these will be the leading technology in 10}20 yr. The production technology for thin-"lm modules differs signi"cantly from that for crystalline silicon modules. In general, thin-"lm modules are made by depositing a thin (0.5}10 μm) layer of semiconductor material on a substrate (usually a glass plate). The thin-"lm deposition can be perfomed with a variety of di!erent techniques, among which are chemical vapour deposition, evaporation, electrolytic deposition and chemical bath deposition. Depending on the selected deposition technique the material properties, material utilization rate and the energy consumption for the deposition process will vary. Processes employing elevated temperatures and/or vacuum conditions will generally have a higher energy consumption per m processed substrate area. Contact layers are also deposited with chemical vapour deposition (transparent contacts) and evaporation (back contacts). Between the subsequent layer deposition steps the individual solar cells are de"ned by removing some of the previously deposited material with laser scribing. When the processing of the solar cells has been completed the module is "nally encapsulated with a second glass plate or with a polymer "lm. Depending on the intended application and on the manufacturer the module may be left frameless or it may be equipped with an aluminium or polymer frame. The main di!erence with crystalline silicon technology is that with thin-"lm technology in each process step the entire module area of 0.5}1m is processed. In crystalline silicon technology, on the other hand, individual wafers of 100}200 cm need to be cut, processed and subsequently combined into a module. Presently, thin-"lm modules have considerably lower conversion e$ciencies than c-si modules, with values in the range of 5}8%. Today amorphous silicon (a-si) is the leading technology in the market and Table 2 shows the breakdown of energy requirements for this type of module, again expressed as MJ per m module area (Alsema, 1998). In the table we can see that the material for the actual solar cell requires relatively little energy. This is of course because of the small cell thickness in amorphous silicon and other thin-"lm technologies. For the rest the energy requirements are divided about equally between module Table 2 Breakdown of the energy requirement of an amorphous silicon thin- "lm module using present-day production technology (glass}glass encapsulation; in MJ of primary energy) Process Cell material 50 Module encapsulation material 350 Cell/module processing 400 Overhead oper. & equipment manuf. 400 Total module (frameless) 1200 Module frame (aluminium) 400 Total module (framed) 1600 Energy requirements (MJ /m module)

5 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999} materials, processing, overhead operations and equipment manufacturing and, "nally, the aluminium frame. Again, frame materials and dimensions may di!er greatly between manufacturers. Polyurethane frames for example require much less energy. In this case the module encapsulation was assumed to comprise two glass plates because this is most common in present-day thin-"lm technology. Switching to polymer back covers, like in crystalline silicon technology, may reduce the energy requirement by some 150 MJ/m. However, in case of toxic solar cell materials such as CdTe or CuInSe (Fthenakis et al., 1999) this would be less desirable (Alsema et al., 1997). Assuming a conversion e$ciency of 7% we arrive at an energy requirement for present amorphous silicon modules of 17 MJ/Wp without frame and 23 MJ/Wp including an aluminium frame. For other thin-"lm technologies (e.g., CdTe, CuInSe, `organic cella) the energy requirements for processing may be di!erent, depending on the type of deposition processes that are used to lay down the cell materials. We expect that in general these di!erences will not be more than 200 MJ/m (Alsema, 1998), although a recent Japanese study suggests a much higher processing energy for a speci"c CdTe deposition process (Kato et al., 1999). Furthermore, overhead operations may require more energy, for example for more extensive waste treatment (cadmium recovery). Nonetheless, we estimate that the energy requirement of most thin-"lm modules will fall within a range of 1000}1500 MJ/m, excluding the frame Other PV system components The requirements for the Balance-of-System, that is all PV system components apart from the modules, will depend largely on the desired application. In gridconnected PV systems, as we consider here, a dc-to-ac converter, cables and some module support materials will be needed. In autonomous systems a battery for energy storage will be required. Like in economic analyses of PV systems the Balanceof-System cannot be neglected in energy analyses. Therefore, we will shortly analyse the impacts of the inverter and array supports. The energy requirements for the converter and cabling are relatively low (&1 MJ/Wp), but module or array supports, on the other hand, do have a signi"cant impact. Recently, the results of a detailed analysis of the primary energy content of present applications of PV systems in buildings have been published by Frankl (Frankl et al., 1998). This study has considered several Note the di!erence in units: the energy requirements for converter and cabling are expressed per Watt, while material and thus energy requirements for array supports are related to the module area. applications on rooftops and building facades, as well as a 3.3 MWp ground-mounted system in Serre, Italy. The array support materials for the Serre plant required an energy input of more than 1800 MJ/m. On the other hand, for roof-integrated modules the Balance-of-System energy requirements are around 700 MJ/m. 4. Energy balance of grid-connected PV systems We will now look at the energy output of a typical PV system and evaluate the EPBT. We will consider two types of grid-connected PV systems, namely a rooftop system and a large, ground-mounted system and two module technologies. We further assume a system performance ratio of 0.75, which means that 25% of energy losses occur on the system level, e.g. in the inverter, cabling, etc. This loss percentage may seem high, but in fact it is representative even for well-designed PV systems today (Erge et al., 1998). As before the conversion e$ciency of the conventional electricity supply system is set at 35%. Finally, we distinguish three irradiation levels, namely: high irradiation (2200 kwh/m yr), as found in the southwestern USA and Sahara; medium irradiation (1700 kwh/m yr), as found in large parts of the USA and southern Europe; low irradiation (1100 kwh/m yr), as found in middle Europe (Germany). Given these assumptions we evaluate the EPBT as the ratio of the total energy input during the system life cycle and the yearly energy generation during system operation (see Fig. 1). We see in the "gure that the EPBT for present-day systems is 2}3 yr in a sunny climate and increases to 4}6 yr (or more) under less favourable conditions. Also note that the contribution from the supports and the module frames is signi"cant, especially for the groundmounted systems. Regarding thin-"lm technology we can see that, due to their lower e$ciency, the energetic advantages of present a-si modules are partially cancelled by the higher energy requirements for frames and supports. An EPBT of 2}6 yr may seem rather long, but in view of the expected lifetime of PV systems of 25}30 yr there is still a signi"cant net production of energy. The net energy yield of these systems, assuming a 30 yr life time, would be 10}14 in sunny regions and 5}7 under low irradiation levels. A "rst conclusion from these energy balance considerations can be that grid-connected PV systems already have a signi"cant potential for reducing fossil energy consumption presently, although it may take a few years

6 1004 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999}1010 Fig. 1. The energy pay-back time (in years) for present-day PV applications. Two di!erent module technologies (multicrystalline silicon and thin "lm * amorphous silicon) and two types of installation (ground-mounted and roof-integrated) are distinguished. of system operation before these savings can actually be cashed in. Nonetheless, it would be preferable if the energy balance could be improved further. Therefore, we will investigate in the next section what prospects exist for a further reduction of energy requirements for PV systems. 5. Outlook for future PV systems In this section we will investigate which improvements in photovoltaic technology may contribute to an improvement of the energy balance of PV systems. The general themes that will be discussed are: material e$ciency, energy e$ciency, new processes, and enhanced module performance. First we will look at crystalline silicon technology Crystalline silicon modules We have seen that in crystalline silicon technology the silicon wafer forms the major contributor to the energy requirements of the module. This o!ers three avenues for reduction of energy inputs, as given below in decreasing order of probability. (1) Reduced wafer thickness: This possibility is already explored by the PV industry as it o!ers signi"cant cost advantages. Wafer thickness reductions from the present-day 300}350 μm down to 200 μm or even 150 μm seem possible, so that the silicon requirements can go down with 30}40% (wafer sawing losses will increase slightly). (2) Other wafer production methods: When blocks of crystalline silicon are sawed into thin wafers about 50% of the material is lost. Furthermore, some 40% is lost in the preceding process of block casting. So only 30% of the electronic-grade silicon ends up in the wafer! As there are no clear possibilities for reduction of the sawing losses, novel methods to produce wafers directly from molten silicon or from silicon powder have been investigated since the 1970s. Although the "rst approach is now in use in at least one commercial production line (edge-de"ned "lm growth), the actual yields of this process are not known. Major bottlenecks in these alternative processes are wafer quality and production speed. However, if such direct wafer production processes become commercially attractive they may lead to signi"cant reductions in silicon requirements, possibly in the order of 40}60%.

7 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999} (3) Other sources for high-purity silicon feedstock: Table 2 shows us that the high-purity silicon used for solar cells is a very energy-intensive material (&1100 MJ/kg). Presently, the PV industry relies fully on silicon which is rejected by the electronics industry because of insu$cient quality (o!-spec material). With the continued growth of the PV market the supply of o!-spec silicon will soon become insuf- "cient so that other feedstock sources will have to be developed. Because standard electronic-grade silicon is too expensive for PV applications, dedicated silicon puri"cation routes will be needed. Not much is known about the energy requirements of such `solar-gradea silicon processes but it seems likely that they are lower than that of the standard process which now produces to the speci"cations of the electronics industry. Regarding the actual cell and module production we can expect additional energy reductions from: (1) using frameless modules; (2) scaling up of production plants, resulting in more e$cient processing, lower overhead, less equipment energy. Summarizing, we expect that future multicrystalline silicon production technology may achieve a reduction in energy requirements to around 2600 MJ/m, assuming innovations like a dedicated silicon feedstock production for PV applications, improved casting methods and reduced silicon requirements (Phylipsen and Alsema, 1995; Frankl et al., 1998; Kato et al., 1998). This kind of technology will probably become available in the next 10 yr. If we further assume a future module e$ciency of 15% (cf. Table 3) we obtain energy requirements of 17 MJ/Wp for multicrystalline-si technology around Modules based on monocrystalline silicon modules will probably remain somewhat more energy intensive, at 20 MJ/Wp (Frankl et al., 1998; Alsema et al., 1998). When we look at the situation beyond 2010, then it seems di$cult to achieve major energy reductions for wafer-based silicon technology. An energy-e$ciency improvement in the production process of 1% per year, as is often found for established production technologies (Beer, 1998), seems therefore a reasonable assumption. Further improvements in the energy requirement per Wp will have to be achieved by improving module e$ciency (while not increasing energy consumption). If we assume that in 2020 the e$ciency of commercial mc-si cells has been increased to 20% (the current record for small-area mc-si cells), then module e$ciency would be about 17% Under our assumptions 2 kg of Si feedstock is needed per m module with the present-day technology. Table 3 Assumptions for module e$ciency development for di!erent cell technologies (in %) Multicrystalline silicon Thin "lm and thus the lowest conceivable energy requirement for Si wafer technology might be 13 MJ/Wp Thin-xlm modules Because the encapsulation materials and the processing are the main contributors to the energy input, the prospects for future reduction of the energy requirement are less clearly identi"able as was the case with c-si technology. A modest reduction, in the range of 10}20%, may be expected in the production of glass and other encapsulation materials. It is not clear whether displacement of the glass cover by a transparent polymer will lead to a lower energy requirement. Other, existing trends which may contribute to a lower energy requirement are: 1. frameless modules; 2. thinner cells with a reduced processing time and thus a reduction in the processing energy and in the energy for equipment manufacturing; 3. an increase of production scale, leading to lower processing energy, lower equipment energy and lower overhead energy. By these improvements we expect the energy requirement of thin-"lm modules to decrease by some 30%, to 900 MJ/m, in the next 10 yr (Alsema, 1996; Kato et al., 1998). If concurrently the module e$ciency can be increased to 10%, the energy requirement on a Wp basis may reach the 9 MJ level for frameless modules. If we try to make projections beyond 2010 we can note that further reductions in the energy requirement below 900 MJ/m do not seem very probable. Like before we assume a generic 1% per year energy-e$ciency improvement in the production process. Only if completely novel module encapsulation techniques are developed which require much less (energyintensive) material we may obtain a more signi"cant improvement. Furthermore, new methods for cell deposition, which require less processing and less overhead operations, might help to reduce the energy input of thin-"lm modules. Of course, module e$ciency increases will directly improve the energy input per Wp (if energy input per m is constant). In this respect signi"cant variations may

8 1006 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999}1010 occur between di!erent types of thin "lms. Moreover, signi"cant e$ciency improvements for thin "lm technology may be achievable. For instance, if we assume a 15% module e$ciency for 2020, the energy requirement per Wp may come down to 5}6MJ Other system components We expect that material and thus energy requirements for the supports in large ground-mounted PV systems cannot be reduced much below the present value of 1850 MJ/m. The Serre PV plant that was considered above is a state-of-the-art design requiring much less Balance-of-System materials than similar European plants (Frankl et al., 1998). Therefore, it seems reasonable to take this "gure as representative for future plants of this type, except for a 1% yearly improvement in the energy e$ciency of the production processes of common materials. When analysing present roof-top systems Frankl found that these systems generally had a considerable scope for improvement, among others by reduced material requirements and increased use of secondary aluminium. Therefore, Frankl asserts that future roof-top systems will have an energy requirement for the support structures of around 500 MJ/m. Table 4 summarizes our expectations for Balance-of- System components and for module frames Energy pay-back time of future systems With the estimates given above on the future energy requirements of modules and Balance-of-System components and using the assumptions on module e$ciency developments we can determine the expected EPBT for these future technologies. In Fig. 2 these results are given using the same assumptions on system performance ratio and electricity supply e$ciency as before, but for only one irradiation level (1700 kwh/m yr). The data for the present situation are also depicted. We can see that according to our energy input estimates and based on our e$ciency assumptions the EPBT values of future PV technology could improve signi"- cantly, resulting in less than 1.5 yr for roof-top systems and less than 2 yr for ground-mounted systems. We also Table 4 Energy requirements for Balance-of-System components and module frame Unit Module frame (Al) MJ/m Array support * central plant MJ/m Array support * roof integrated MJ/m Inverter (3 kw) MJ/W see that thin-"lm modules may gain a signi"cant advantage over c-si technology when considering roof-top applications, if the module e$ciency can be increased to 10% or higher. Furthermore, it is quite clear that rooftop installations have a much better potential for low EPBT than ground-mounted installations. Our conclusion is that we see good possibilities for future developments in PV technology that result in decreasing energy requirements in component production and thus lead to an increasingly higher potential for fossil energy displacement. Below we will discuss to what extent these technology developments will take place autonomously or will be driven by cost reduction goals, or that they will require speci"c policy actions. First, however, we want to take a look at the emissions of PV systems and compare them with other energy technologies in order to gain insight into the mitigation potential of PV technology. 6. The potential for CO 2 mitigation For a start we can remark that, for PV systems, the EPBT is also a quite good indicator of the mitigation potential because generally more than 90% of the greenhouse gas emissions during the PV system life cycle are caused by energy use (Dones and Frischknecht, 1998). Emissions not related to energy use are only found in steel and aluminium production (for the supports and frames) and in silica reduction (for silicon solar cells). To obtain the emissions due to the production of a PV system we have to multiply all energy and material inputs with their corresponding emission factors. This requires a detailed life-cycle assessment of greenhouse gas emissions from solar cell manufacturing (and other life-cycle stages) and from the life cycle of Balanceof-System materials. Such an analysis goes beyond the scope of this paper. Here we will make a rough estimate of emissions by: (1) considering only energy-related emissions during module production, (2) by evaluating all energy inputs in module production as electrical energy and (3) considering only the aluminium supports employed in roof-integrated systems for the Balance-of-System components. We estimate the overall error of these approximations at maximally 20%. Results from such a full life-cycle assessment, but based on less recent energy requirement estimates, can be found in Dones and Frischknecht (1998). Non-energy emissions are 0.1 kg/wp (silica reduction) or less. Primary energy use for electricity is about 90% of total primary energy consumption for crystalline silicon modules. For thin-"lm modules this share is about 70%, but the remaining 30% is used in glass production, where by chance the emission is similar as our assumption for electricity production (0.055 kg per MJ of used energy).

9 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999} Fig. 2. The energy pay-back time (in years) for two representative PV applications, both for present-day and for future (2010 and 2020) technology. The "gures in parentheses denote the assumed module e$ciencies for each option. Note that actual pay-back times will vary with irradiation and system performance. We now arrive at a quite important point regarding the emissions of PV systems. Di!erently than with most other energy technologies the emissions of PV occur almost entirely during system manufacturing and not during system operation. As a consequence the emissions of PV are determined very strongly by the `manufacturing environmenta and more speci"cally the fuel mix of the electricity supply system that is employed during the PV system production. In other words the emission of a PV system will depend on the emission factor of the local utility system. In this analysis we have assumed the present-day fuel mix of continental Western Europe (UCPTE region), where about 50% of the electricity is produced by nuclear and hydro-electric plants, as well as 20% by coal-, 10% by oil-, and 10% by gas-"red plants. For this utility system the emission factor is presently about 0.57 kg per kwh produced electricity (&0.055 kg/ MJprim (Suter and Frischknecht, 1996)). Although the fuel mix is likely to change in the future, the emission factor will probably change with less than 7% (Dones and Frischknecht, 1998) up to For simplicity we will therefore take this factor as constant. Two other parameters, which may not be overlooked when evaluating emissions of PV, are, of course, the irradiation at the site of PV system installation (cf. Section 3) and the assumed system lifetime. As before we have assumed a medium high irradiation of 1700 kwh/m yr and a system lifetime of 30 yr. Note that because of the three parameters mentioned above emission estimates for PV can vary signi"cantly between di!erent studies. In Fig. 3 we have displayed the emissions per kwh of supplied electricity for grid-connected rooftop PV systems. For comparison a number of conventional power generation technologies are also depicted (all status 1999). From a recent IEA study (IEA, 1998) we have furthermore included estimates for wind turbines and biomass gasi"cation technology as well as their estimate for present-day multicrystalline-silicon PV technology in a roof-top application. The results show that according to our estimates the emissions for present PV technology are in the range of 50}60 g/kwh which is considerably lower than for fossil-fuel plants but higher than for the two competing renewable energy technologies. Our estimate for mc-si technology seems somewhat lower than that from the IEA study, but if we consider that the latter probably has assumed a 30% higher emission factor for the utility system, the di!erence becomes negligible (which is to be expected as the respective energy estimates are also comparable). With improving technology PV-related emissions may become signi"cantly lower, around Corrected for the di!erent assumptions on irradiation and system lifetime. The assumed fuel mix or the emission factor of the electricity supply system is not given exactly (`current Germana) but their emission factor is presumably some 30% higher than ours. We did not correct for this di!erence. For biomass energy it is even argued that the emission factor is e!ectively zero because parts of the biomass crop are not harvested and stay in the ground (trunks, etc.), thus forming a sink which compensates for the energy consumption in harvesting and transport.

10 1008 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999}1010 Fig. 3. emission for grid-connected roof-top PV systems now and expected in the future. For comparison we show emission data for a number of competing energy systems (coal, gas, nuclear data from Suter and Frischknecht (1996); wind and biomass energy estimates from a recent IEA study (IEA, 1998)). The PV technology estimate from the same IEA study is also shown. Note that actual emissions for PV will vary with irradiation and system performance. 20}30 g/kwh in the near future, or even 10}20 g/kwh on the longer term. Only in the last case PV systems come into same range as current wind, biomass and nuclear energy. In conclusion, we can say that PV systems can supply energy at a considerably lower emission rate than current fossil energy technologies, but on the other hand PV is slightly at a disadvantage when compared to biomass and wind energy. 7. Policy implications We will now discuss the policy implications of our analysis results. First we will focus on the implications regarding photovoltaic R&D policy and policies aimed at market stimulation for PV systems. After that we look into the implications for climate change policy. We have shown above that grid-connected PV systems currently have an EPBT that is well below their expected lifetime. So in this respect there is no reason to abandon investments in PV technology or photovoltaic R&D. At the same time, however, we have to concede that improvements in the energy balance are desirable if we want to promote PV as one of the major options for mitigation. Such improvements can be obtained by three types of measures: 1. reduction of energy consumption in PV system production; 2. enhancement of PV system performance (i.e. higher module e$ciency); 3. increase of system lifetime. While the last two measures will be strived for anyway, because they reduce PV energy costs too, this is not necessarily true for the "rst type of measures. Therefore, we look into them in some more detail to see whether policy actions are needed here. In crystalline silicon solar cell technology there is already a trend towards thinner wafers, because it can reduce costs. The energy balance will surely pro"t from this trend. The R&D into new wafer production techniques that can reduce the material losses inherent in sawing is also driven by costs. And if they do not become economically attractive there is no reason to pursue them further. In the third place we have seen that the high energy intensity of electronic-grade silicon is a major bottleneck. As the demand for silicon solar cells grows it seems probable that some kind of adapted or fully new silicon puri"cation process will become available which delivers material to the speci"cations of the solar cell industry. However, it is less sure whether such a process will also have signi"cantly reduced energy demands. So in our view this should be a point of attention in R&D funding programmes. In thin-"lm technology less clearly identi"able options for energy reduction seem to exist. As a major share of the energy was consumed during cell and module processing, the only way seems to be to increase energy

11 E.A. Alsema, E. Nieuwlaar / Energy Policy 28 (2000) 999} e$ciency and energy awareness within this industry itself. Because energy is a relatively minor cost factor and because production technologies in this sector are changing very rapidly the attention for energy consumption in the solar cell plant itself seems to be relatively small. So this seems like a point for policy actions too. Finally, we want to stress the importance of good system design. Modules themselves and module supports should be designed in such a way that the material and energy requirements are low, without compromising system lifetimes. A good example seems to be the deployment of frameless modules thus avoiding the use of energy-intensive aluminium pro"les. At this point the PV industry seems a bit conservative and reluctant to abandon proven methods. Also in the design of support structures excessively heavy structures may be found and very often ample use is made of convenient aluminium pro- "les. And if the latter are really indispensable why not use more secondary (recycled) aluminium instead of primary? With respect to support structures it is also very important to note that roof integration o!ers signi"cant advantages over ground-mounted structures. Apart from these policy issues aimed at improving the energy and record for PV technology we would also like to discuss the implications of our results for climate change policy. In other words what conclusions can be drawn regarding the (future) role of PV technology in abating climate change? Because the EPBT of current PV systems lies between 2 and 6 yr there will always be a time lag of some years before deployment of a PV system will result in a reduction of fossil energy consumption and thus a lowering of emissions. Note that the actual length of this time lag is determined not only by the EPBT value, but also by ratio of the emission factors of the electric utility supplying the PV production plant, on the one hand, and the power plant(s) which are being displaced by the PV energy, on the other hand. Unfortunately, this time lag is neglected in almost all models evaluating the emission reduction by deployment of PV. If we furthermore combine the time lag e!ect with the fact that only a limited PV capacity can realistically be installed up to 2010 (e.g., in view of production capacity limitations) we have to conclude that the contribution of PV systems to mitigation will be very limited in the next decade. In other words, deployment of PV technology will not help much to meet the Kyoto targets. On the other hand, our analysis also shows that PV technology does certainly o!er a large potential for mitigation when looking beyond And after all, climate change and mitigation are issues which will So ¹ "EPBT E /E. Note that this is only a "rstorder approximation. probably remain with us long after The reasonably low emission rate, the large technical potential and the good environmental pro"le (Fthenakis and Moskowitz, 1995; Alsema, 1996; Nieuwlaar and Alsema, 1997, 1998; IEA, 1998) of PV technology make it one of the most interesting candidates for a long-term sustainable energy supply. 8. Conclusions In this paper we have reviewed the energy viability of photovoltaic energy technology, that is the question whether photovoltaic (PV) systems can generate su$cient energy output in comparison with the energy input required during production of the system components. We evaluated the energy viability mainly in terms of the energy pay-back time (EPBT). For a grid-connected PV system under a medium-high irradiation level of 1700 kwh/m yr the EPBT is presently 2.5}3 yr for roof-top systems and almost 4 yr for large, groundmounted systems. The share of module frames and module supports in this "gure is quite signi"cant. Under other climatic conditions the EPBT values will be inversely proportional to the irradiation. Furthermore, we have shown that there are good prospects that in the coming 10 yr the EPBT of roof-top systems will decrease to less than 2 yr, if certain improvements, both in production technology and in module performance are realized. For modules based on crystalline silicon one of the requirements would be a dedicated silicon puri"cation process with substantially lower energy consumption, while for thin-"lm technology mainly an improved module e$ciency would be necessary. Roof-top-type systems will maintain their advantage over ground-mounted systems. emissions from roof-top PV systems were calculated assuming that the PV production facility is located in Western Europe. It was found that the speci"c emission at present is about 50}60 g/kwh, which is considerably lower than the emission from existing fossil-fuel electricity plants (400}1000 g/kwh). On the other hand, these emission values for PV systems are higher than what is estimated for two major competing renewable technologies, wind and biomass ((20 g/kwh). Given the expected technology improvements the speci"c emission from PV could go down to 20}30 g/kwh in the next 10 yr and perhaps even lower after However, in some areas policy actions may be needed to realize this, for example by stimulating energy-e$cient production processes for PV modules and by promoting system designs which require less materials and energy. We have pointed out two consequences of the relatively high energy input during PV system manufacturing, a situation that is rather unique for energy technologies.