Energieonderzoek

Learning in renewable energy technology

Summary and Conclusions


1. Introduction

Renewable energy sources are not limited by finite fuel reserves. They have a large technical potential to contribute to global energy needs and this potential is geographically more evenly distributed than fossil fuel reserves are. In general, their application also has lower external (e.g. environmental) costs than the present use of fossil fuels. These characteristics have been key drivers for the Dutch government to set ambitious targets for the production of electricity from renewable sources in 2010 and 2020: a contribution of respectively 9% and 17% to the gross domestic electricity consumption. Yet, the current contribution is only 3.3% and it is uncertain whether these targets will be reached. The efforts to accelerate the implementation of renewable electricity can be hindered by several barriers, such as technical, economic, social and institutional barriers. A major barrier to a large-scale diffusion of renewable energy technologies are the electricity production costs. These costs have been re duced in the past few decades for a number of renewables and are expected to decline in the future too due to technological learning. A frequently used approach to quantify and evaluate past cost reductions and to project potential future cost reductions is the experience curve approach. The experience curve describes the cost development of a product or technology as a function of cumulative production. The approach has been applied to the development of renewable electricity technologies frequently, especially onshore wind turbines and solar photovoltaic energy modules and systems. However, many methodological questions when using this approach have not been answered yet, such as the appropriate geographical boundaries of learning systems, the use of a so-called compound learning system approach (in which the main learning system is disaggregated, and the cost development of components is investigated), and the question whether the slope of these curves is, or remains constant or not. A question is also whe ther the experience curve approach is suitable to describe the cost development of biomass fuelled plants. Given the range of application of experience curves, especially for policy advice and in energy models, further insights are required on how to deal with these issues.

2. Thesis objectives and research questions

The main objectives of this thesis are:

  • To investigate technological change and cost reduction for a number of renewable electricity technologies by means of the experience curve approach,
  • To address related methodological issues in the experience curve approach, and, based on these insights,
  • To analyze the implications for achieving the Dutch renewable electricity targets for the year 2020 within a European context.

In order to meet these objectives, a number of research questions have been formulated:

  1. What are the most promising renewable electricity technologies for the Netherlands until 2020 under different technological, economic and environmental conditions?
  2. To what extent is the current use of the experience curve approach to investigate renewable energy technology development sound, what are differences in the utilization of this approach and what are possible pitfalls?
  3. How can the experience curve approach be used to describe the potential development of partially new energy technologies, such as offshore wind energy? Is it possible to describe biomass fuel supply chains with experience curves? What are the possibilities and limits of the experience curve approach when describing non-modular technologies such as large (biomass) energy plants?
  4. What are the main learning mechanisms behind the cost reduction of the investigated technologies?
  5. How can differences in the technological progress of renewable electricity options influence the market diffusion of renewable electricity technologies, and what implications can varying technological development and policy have on the implementation of renewable electricity technologies in the Netherlands?

The development of different renewable energy technologies is investigated by means of some case studies. The possible effects of varying technological development in combination with different policy backgrounds are illustrated for the Netherlands. The thesis focuses mainly on the development of investment costs and electricity production costs. Possible additional costs of intermittent renewable electricity sources (such as storage, backup-capacity or grid fortification) with advanced penetration are not investigated, although these issues may be important on the longer term (after 2020).

3. Summary of the findings

It is uncertain whether and under which conditions the Dutch policy goal of realizing a contribution of 17% from renewable sources to the domestic electricity demand in 2020 (i.e. 18-24 TWh) can be achieved. Chapter 2 explores the feasible deployment of renewable electricity production in the Netherlands until 2020 by evaluating different images representing policies and societal preferences. First Dutch policy goals, governmental policy measures and definitions of renewable electricity are discussed. Second, a comparison is made of four studies that analyze the possible development of renewable electricity production in the coming decades. Finally, three images are set up. In each image, the impact of a key factor that influences the maximum realizable potential (economical performance, ecological sustainability and technological progress) is evaluated. Results show that for the onshore wind potential, environmental criteria and available space are the main limiting parameters. With a realization of 5-7.5 TWh electricity production a year in all images, it is a relatively robust option in all three images. Wind offshore also obtains a significant share in all three images. The largest uncertainties of this yet unproven technology are the successful technological development, the possibility to build plants within the 12 mile zone (which is economically attractive, but less desirable from an environmental point of view), and the maximum installation rate that can be achieved until 2020. With regard to biomass, two different technologies are evident: different forms of co-firing biomass as the most economical option, or large-scale (stand-alone) gasification plants as the most efficient technology. In all images, either large-scale co-firing in coal plants and natural gas combined cycle (NGCC) plants or biomass integrated gasification / combined cycle (BIG/CC) plants contribute substantially to the total renewable electricity production. In the image with high technological progress and implementation rates, an annual production of 42 TWh may be achieved in 2020, mainly due to the large penetration of offshore wind farms and BIG/CC plants. Under stringent economical or ecological criteria, about 25 TWh may be reached. The three scenarios are not ‘best guess’ scenarios, and no integration of them was carried out. When only the robust options (i.e. options present in all three scenarios) are considered, 9-22 TWh can be realized. The analysis illustrates the importance of taking the different key factors mentioned influencing implementation into account. Doing so allows for identification of robust and less robust technological options.

Chapter 2 revealed that the technological development of various renewable electricity options and the associated reduction in production costs can have a major influence on their market diffusion. Therefore, in the following chapters, the past and potential future technological development of onshore wind farms (chapter 3), offshore wind farms (chapter 4) and different biomass energy systems (chapter 5 and 6) are investigated. As the development of these technologies has occurred on in many different countries, several international case studies are carried out.

In chapter 3, technological development and cost of wind farms and the production costs of wind electricity are investigated, using the experience curve approach. Experience curves of wind turbines are generally based on data describing the development of national markets, which cause a number of problems when applied for global assessments. To analyze global wind energy price developments more adequately, a global experience curve is composed. First, underlying factors for past and potential future price reductions of wind turbines are analyzed. Also possible implications and pitfalls when applying the experience curve approach are assessed. An explicitly investigated issue are the geographical boundaries of learning systems. Within this frame, the price development of wind turbines and wind farms and the effects of support policy in Germany are evaluated. It is concluded, that the German support policy has caused prices to remain stable since 1995, making the German data unsuitable to determine the gener al speed of technological learning of wind turbines. Based on these insights, an approach is presented to establish a global experience curve and thus to determine a global progress ratio (PR) for the investment costs of wind farms, based on wind farm price data from the UK and Spain, which are deemed to follow production costs more closely. Results show that global PRs for wind farms may lie between 77-85% (with an average of 81%), which is significantly more optimistic than PRs applied in most current scenario studies (based on the construction of national experience curves) and integrated assessment models. While the findings are based on a limited amount of data, they may indicate faster price reduction opportunities than so far assumed.

Chapter 4 investigates the potential cost reduction prospects of offshore wind farms. The economics of wind farms offshore are presently less favorable than for onshore wind farms. Consequently there is a strong need for significant cost reductions in order to become competitive. About 70% of the electricity cost of offshore wind farms is determined by the initial investment costs, which mainly consist of the wind turbines, foundations, internal and external grid-connections and installation. Possible cost reductions until 2020 are explored for each of these components. Technological developments and cost reduction trends in both the offshore and onshore wind sector are analyzed. Information is also taken from offshore oil and gas sector and from the experience with high-voltage submarine transmission of electricity. Where possible, cost reduction trends are quantified using the experience curve concept, or otherwise based on expert judgments. Main drivers for cost reduction appear to be (a) design improve ments and upscaling of wind turbines, (b) the continuing growth of onshore wind capacity, and (c) the development and high utilization rates of purpose-built installation vessels. Other factors are: reduction of steel prices, technological development of HVDC converter stations and cables, standardization of turbine and foundation design, and economies of scale for the wind turbine production. It is concluded that it is possible to use the experience curve approach for offshore wind farms, by using data from related industry sectors. Under different growth scenarios, investment costs of offshore wind farms may decline from 1600-1700 €/kW in 2001 to 980-1300 €/kW in 2020. Assuming an identical relative decline of annual O&M costs, the levelized electricity production costs may be reduced by 25-39% compared to current costs. The analysis also reveals, that under the presumptions of mutual learning and cost reductions with other technologies, only 15% of all cost reductions are related directly to the instal lation of offshore wind farms alone, while 80% are partially depending on the further development of involved technologies (onshore wind turbines, HVDC converter stations and submarine cables, and offshore steel and concrete foundations).

In chapter 5, the focus switches to the development of biomass energy systems. An important part of bioenergy systems is the fuel supply. With its increasing use for heat and electricity production, the production costs of Primary Forest Fuel (PFF: branches, tops, small trees and un-merchantable wood left in the forest after the cleaning, thinning or final felling of forest stands) have declined over the last three decades in Sweden. The aims of chapter 5 are to quantify cost reductions of PFF production as achieved in Sweden over time, to identify underlying reasons for these reductions, and to determine whether the experience curve concept can be used to describe this cost reduction trend. Also the suitability of this concept to project future PFF cost reductions in Sweden and in other countries is explored. The analysis was done using average national PFF price data (as a proxy for production costs), a number of production cost studies and data on annual Swedish production volumes. Results show that main cost reductions were achieved in forwarding and chipping of PFF, largely due to learning-by-doing, improved equipment and changes in organization. The price for wood fuel chips does follow an experience curve from 1975-2003 (with over nine cumulative doublings). The PR is calculated at 87%. However, given the uncertainty in data on PFF price and annual production volum es, the PR may range between 85% and 88%. It is concluded that in combination with the available supply potential of PFF and taking into account the results of a bottom-up assessment of cost reduction opportunities, the experience curve can be a valuable tool for assessing future PFF production cost development in Sweden. A methodological issue that needs to be further explored is how learning took place between Sweden and other countries, especially with Finland, and how the development of technology and PFF production in these countries should be combined with the Swedish experiences. This would allow the utilization of the experience curve concept to estimate cost developments also in other countries with a large potential to supply PFF, but with less developed PFF supply systems. It should also be investigated, how local knowledge and technology can be transferred to these countries, which is likely to be crucial to achieve low PFF costs.

The main goal of chapter 6 is to determine whether cost reductions in different bioenergy systems can be quantified using the experience curve approach, and how specific issues (arising from the complexity of biomass energy systems) can be addressed. This is pursued by case studies on biomass-fuelled combined heat and power (CHP) plants in Sweden, global development of fluidized bed boilers, the results from chapter five, and the development of biogas plants in Denmark. As secondary goal, the aim is to identify learning mechanisms behind technology development and cost reduction for the biomass energy systems investigated. The case studies reveal large difficulties to devise empirical experience curves for investment costs of biomass fuelled power plants. To some extent, this is due to general lack of (detailed) data, but mainly because of varying plant costs due to differences in scale, fuel type, plant layout, region, etc. Only in a few cases, some meaningful trends were found. For plants utilizing fluid ized bed boilers, PRs for the entire plant lie between 90-93% (a range which has also been found for other large plants using a number of fuels), but the costs for the boiler section alone may decline much faster. The experience curve approach delivers better results, when the production costs of the final energy carrier are analyzed. Electricity costs from biomass-fuelled CHP-plants yield PRs of 91-92%. The experience curve for biogas production costs displays a PR of 85% from 1984 to the beginning of the 1990s, and levels afterwards to approximately 100% until the year 2002. For technologies developed on a local level (e.g. biogas plants), learning-by-using and learning-by-interacting are important learning mechanism, while for CHP plants utilizing fluidized bed boilers, upscaling is probably one of the main mechanisms behind cost reductions.

The objective of chapter 7 is to examine the consequences of differences in technological developments on the market diffusion of specific renewable electricity technologies in the EU-25 until 2020, using a market simulation model (ADMIRE REBUS) developed by the Energy research Centre of the Netherlands (ECN). For the main analysis, it is assumed that from 2012 a harmonized trading system will be implemented, and a target of 24% renewable electricity (RES-E) in 2020 is set and met. The results from the previous chapters were used to set up an optimistic and a pessimistic endogenous technological learning scenario. It was found that the diffusion of onshore wind energy is relatively robust, i.e. independent of technological development, but the diffusion rates of offshore wind energy and biomass gasification greatly depend on assumptions about their technological development. Competition between these two options and (conventional) biomass combustion options largely determines the overall costs of electrici ty from renewables and the choice of technologies for the individual member countries. In the optimistic scenario, in 2020 the market price for RES-E is 1 €ct/kWh lower than in the pessimistic scenario (7 vs. 8 €ct/kWh). As a result, total expenditures for RES-E market stimulation are 30% lower in the optimistic scenario. For comparison, also the impact of continuing present support policies until 2020 was evaluated assuming no international trade of RES-E certificates. As Member states then have to achieve their target by exploiting their own potentials only, a number of member states have to utilize their offshore wind potential, making the diffusion of offshore wind electricity production much less dependent on both the rate of technological development and the competition from biomass options, compared to the harmonization scenario.

When the results of chapter 2 and chapter 7 are compared, chapter 2 shows 9-22 TWh as robust potentials under different assumptions. In chapter 7, the analysis shows that under the continuation of present policies, likely 14-16 TWh is realized, depending on the rate of technological development. Clearly, if the Netherlands are going to pursue ambitious renewable energy targets, offshore wind energy use offers the largest potential if only domestic sources are considered and no biomass is imported. However, even under the continuation of the current policy measures, and the optimistic technological learning scenario, in 2020 only the production of 7 TWh is realized. This corresponds to an installed wind turbine capacity of approximately 2200 MW, which is far less than the current governmental target of 6000 MW for 2020. Overall, even under conditions of optimistic technological learning, the target of 17% contribution from renewable sources in 2020 (i.e. 17-24 TWh, depending on growth in electricity demand and other factors) is unlikely to be met. Under the harmonization scenario, a similar domestic production of electricity from onshore wind, offshore wind and biomass occurs, but an additional 2-4 TWh are imported (in the form of certificates), leading to an overall realization of about 19 TWh. Thus, under the harmonization scenario, the Dutch 17% target is more likely to be achieved, as elsewhere in the European Union sufficient other renewable energy can be realized. Remarkably, under the optimistic learning scenario, the Netherlands import more certificates, as more low-cost potential can be exploited in other countries of the European Union.
It is emphasized that the numbers presented here are no forecast, but merely possible developments under the conditions assumed in chapters 2 and 7. For example, as pointed out in chapter 2 and 7, the availability of cheap and sustainable biomass from abroad may play an important role for the overall biomass generated electricity potential, and could increase the overall potential of biomass use in the Netherlands.

4. Methodological lessons

The investigations on the selection of geographical boundaries for learning systems shows that for technologies deployed all over the world (such as onshore wind farms or PV modules), the most appropriate system boundaries are on a global level. However, for technologies, which have been developed on a more local level (such as small scale biogas plants), and no significant exchange of experiences with other learning systems, an analysis on a local or national level may be most appropriate. A major conclusion is that sufficient attention has to be given to determine the appropriate boundaries of the learning system in order to determine the correct PR of a learning curve, as the examples of national onshore wind farms in chapter 3 and fuel supply chains in Finland in chapter 6 illustrate. In addition, sometimes also mixtures occur, as parts of a technology may learn on a local level whereas other parts may learn on a global level, as was shown for onshore wind farms and for biomass fuel chains.

The approach of compound learning systems was used both for offshore wind farms and for biomass energy systems, though with different aims. First, for offshore wind farms, the possible technological development of turbines, foundations, grid-connection and installation was determined by analyzing the PRs of related technologies such as onshore wind turbines, offshore foundations and submarine electricity cables. This was necessary, because only a handful of offshore wind farms have been built so far. The analysis shows, that this approach is feasible, even though only time will tell whether the presented cost reduction potentials will actually be realized. Second, for biomass energy systems, the question was mainly whether or not the experience curve approach was suitable at all to describe the cost development of (different parts of) biomass energy systems. Therefore, a differentiation between the investment costs, the fuel costs and the O&M costs was performed in order to investigate the different cost r eduction trends and to determine the suitability of the experience curve approach for the different sub-learning systems. As was shown in chapters 5 and 6, the experience curve approach can be applied to fuel supply chains and to the cost of the final energy carrier (i.e. electricity and biogas) yielding trend lines with quite satisfactory correlations. However, the experience curve approach appears to be less suitable for measuring the decline of biomass plant investment cost. Further studies regarding the applicability of the experience curve approach to evaluated the progress made in biomass fuel supply chains are recommended. This may result in an appropriate method to analyze cost developments of new biomass energy chains, for example chains based on dedicated crop plantations.

Furthermore, the question remains, whether the PR value may change (and, more specifically, may approach towards 100%) with increasing market diffusion. In two cases (German onshore wind farms and Danish biogas plants), experience curves were found to flatten, the PR value becoming approximately 100%. However, in both cases, it was shown that this was due to changes in the market (i.e. subsidies and fuel shortages), but not as a result of structural changes in technology development. Overall, no indications were found that PRs based on production costs should become less benign, at least as long as the market share of the technology continues to increase.

The difference in use of marginal and average cost was explored for different biomass energy systems. As expected, the marginal cost experience curve lies below the average cost experience curve. However, it also displays the same slope. This may allow for using PRs from experience curves based on historical data of average costs (which may be more readily available) to determine potential further cost reductions of the best available technology.

Related to the issue of geographical learning system boundaries is the matter of correcting for inflation in complex international innovation systems. When data from several countries are involved, the choice of reference currency and the method of using exchange rates can influence the PR significantly, as was shown in chapter 3. While no ideal solution was found for this problem, evaluating the use of several reference currencies may provide insights in the uncertainty of the results.

In addition, the learning mechanisms behind the cost reductions of renewable energy technologies were investigated. Naturally, they differ per technology, but some parallels can be drawn on basis of scale and geographical diffusion. For onshore wind turbines, upscaling has been the most important factor behind cost reductions on a global level in the past. However, learning effects such as the improved siting of wind farms and lower grid connection costs typically occur on a more local level. The development of offshore wind farm has only been possible by the continuous upscaling of onshore wind turbines. Both onshore and offshore wind farms are expected to benefit from the effects of economies of scale with the increasing wind farm size in the future. With regard to biomass technologies, the fluidized bed boiler combustion plants, currently deployed on a global level, has also benefited from gradual upscaling over the past few decades. On the other hand, cost components such as the fuel supply and the ope ration and maintenance of plants largely depend on local knowledge, as the examples of Swedish CHP plants and Danish biogas plants illustrated. In these cases, local learning-by-doing, learning-by-using and learning-by-interacting have been vital mechanisms for the successful development of these technologies. In addition, by determining learning mechanism that may occur in the future (e.g. the mass production of wind turbines), the extrapolation of experience curves for prospective cost development analysis can be supported qualitatively.

5. Implications for the development and market diffusion of renewable energy technologies

In general, the two policy scenarios demonstrate, that the influence of technological learning on the diffusion and cost of renewable energy technologies largely depends on the possibility of competition. A European-wide trade in renewable energy certificates enables the optimal utilization of the cheapest available sources and technologies, and therefore favors technologies with rapidly declining electricity production costs. If trading possibilities are absent, technological learning my have a much lower impact on overall diffusion rates, as shown for offshore wind energy and various biomass energy technologies in chapter 7 of this thesis. Furthermore, as was pointed out in chapter 4 and 7, the development of pilot plants for offshore wind farms and BIG/CC plants requires international cooperation and knowledge exchange, as no single country on its own can build the required number of plants needed to reduce investment costs over a long period of time. On the other hand, as already pointed out, the disse mination of knowledge acquired by the use of technologies is much more limited to the local level, especially for smaller-scale technologies. Therefore, this should achieve specific attention when national policy supports are increased.

The research has shown that in many cases the quantity and quality of data are not sufficient to carry out experience curve based analyses. This is partially due to the confidentiality of data, e.g. on production costs, but also to lack of structured data collection on renewable energy technologies. While this database is available for PV and wind turbines, there is very little data available on most biomass technologies. It is recommended, that this data is collected in a more structured way to enable further analyses. Also in regard to the development of new renewable energy technologies (e.g. offshore wind farms) the data availability of related technologies (such as submarine high voltage electricity cables and steel or concrete foundations) is not optimal.

In the case studies scrutinized in this thesis it was observed that most large-scale technologies are deployed on a global scale, and that therefore also technological learning occurs worldwide. While national research programs to develop new technologies may be very useful in the very first stages of diffusion, in later stages it may prove more fruitful to focus on knowledge exchange and dissemination of local experiences. Within this frame, a EU wide approach to stimulate offshore wind farms and BIG/CC plants may be required, also to achieve the necessary volumes.

Policy makers always have to make a trade off between funding RD&D, funding early market introduction and funding large-scale market diffusion (VROM-raad and AER, 2004). While the optimal funding ratio between these categories is difficult to determine and may depend on the local geographical situation and the technology involved, chapter 7 has shown, that early investments in pilot plants may save large amounts of subsidies later, in the market diffusion phase. However, as this relation is not easily determined beforehand, further research is recommended in this direction.