Skip to main content

System analysis for progress and innovation in energy technologies ('SAPIENT')

Exploitable results

The improved under SAPIENT version of the POLES model has made feasible the construction of endogenous technological progress scenarios. The IEPE team has used the SAPEX scenario (SAPIENT-exogenous scenario involving exogenous forecasts on technology development) as a benchmark for the assessment of the SAPEN (Sapient endogenous) scenario. In the latter case exogenous hypotheses have been replaced by a description of technology cost dynamics that endogenises technological change in the new and renewable technologies and in the electricity modules of the POLES model.
The POLES model is a global sectoral model of the world energy system. It has been developed in the framework of a hierarchical structure of interconnected sub-models at the international, regional and national level. The dynamics of the model are based on a recursive (year by year) simulation process of energy demand and supply with lagged adjustments to prices and a feedback loop through international energy prices. The model is fully operational and can produce detailed long-term (2030) world energy and CO2 emission outlooks with demand, supply and price projections by main region. IEPE has concentrated on work that has allowed the update and enhancement of the POLES Reference Case with a view to specifically serve the needs of the SAPIENT project. Besides the update of the model's entire database up to the years 1997-1998, the following sets of hypotheses has been fully revised: - Demographic and economic growth by POLES region; - Oil and gas resources, - Definition of a full 2030 world CO2 entitlement scenario for the POLES regions, along a "soft landing" path (progressive stabilisation of emissions in developing countries) which was used as a benchmark by the other models in the context of GHG emission abatement scenarios. POLES has been developed to offer detailed energy scenarios for the world in 38 regions for different sets of hypotheses of Government R&D and/or carbon emission constraints.
IER has concentrated on the development of Times to analyse several energy-environment scenarios for Europe. TIMES (The Integrated MARKAL-EFOM System) is a MARKAL class of large scale energy system model, with some additional features as compared to MARKAL which can address some of the recently developed complex energy - environment related issues, like, carbon trading, carbon leakage from one country to another, issues on clean development mechanism etc. It is expected to be the successor of MARKAL. IER has tried to validate TIMES against MARKAL using European energy-material database. Implementation of single factor technology learning curve has been successfully done within the TIMES. The incorporation of two-factor learning curve (TFLC) in TIMES has proved an extremely difficult task, since the model was 8-10 times larger than MARKAL by size. Therefore, only the endogenous single factor technology learning has been introduced in TIMES-WEU.
One of the new features implemented in ERIS by IIASA has been the capability to define R&D expenditures in absolute monetary units. Using this feature allows the model user to not only optimise a given R&D budget, but also to "freely" (without constraints) optimise the R&D support of a technology for which a 2FLC has been specified. This approach naturally does not realistically describe the decision process of EC policy makers, but experimental runs with ERIS in such a hypothetical world enhanced the basic understanding of the model's dynamics and plausible results in this respect boost the confidence in the model's results. IIASA has performed a sensitivity analysis that tests the consequences on optimised R&D expenditures, of assuming the "wrong" learning rates (learning-by-doing rate and learning-by-research rate). In order to focus on testing the dynamics of ERIS in response to changing the parameters of two-factor learning curves, the model was used in an aggregate ("stylised") fashion. Solar photovoltaic and wind electricity generation technologies were used to paraphrase a situation in which one technology has a long way to go before it reaches competitiveness but has a high potential for technological progress, and the other technology being closer to competitiveness, but learns at a slower pace. The sensitivity of the results was then tested with respect to the learning parameters, with focus on the ranges around some reference values. Subsequently, the model was assessed with 2FLC applied to both technologies at the same time, to study how the optimised R&D support of one technology is influenced by the presence of a competitor. The procedure has generated useful results fully analysed in the final report. IIASA has also extended the solution horizon of ERIS to 2100 allowing for the performance of R&D budgeting exercises with a very long-term perspective.
ECS IIASA has also undertaken some estimation of Two Factor Learning Curves using data available by IEPE. Two different versions of the TFLCs were used during the process. One was the original formulation including cumulative R&D expenditures. The other involved using the "knowledge" instead of cumulative R&D. For the latter, the sensitivity of the estimated parameters of TFLCs with respect to initial knowledge stock and past R&D expenditures is also analysed.
In the original MERGE model, technological learning was not considered. Energy technologies instead had fixed characteristics over time. PSI has carried out extensive work developing MERGE-ETL in which endogenous technological learning is applied to eight electric and non-electric energy technologies. Technological learning describes how the specific (investment) cost of a given technology is reduced through the accumulation of knowledge. The latter may have different sources, such as the technology's manufacturing ('learning-by-doing') or research-and-development expenditures (�learning-by-searching�). A learning curve relates then for a given technology its specific cost to one or more factors describing the accumulation of knowledge. The two factor learning curve was implemented, where the specific cost is reduced both as a function of the cumulative capacity and of the cumulative R&D expenditures.
ERIS (Energy Research and Investment Strategy) is a global energy model prototype specified and developed in the context of the predecessor of SAPIENT, TEEM (Joule III project). The original purpose of ERIS was to capture the main mechanisms regarding the endogenous analysis of technological learning under uncertainty and to allow for a consistent cost-benefit analysis of specific policies aiming at technology prioritisation. The original prototype specified by IIASA and coded by NTUA considered a non-linear programming (NLP) and a Mixed Complementarity (MC) formulation of experience curves and it has been restricted to three technologies and one region.. ERIS was extended by PSI to include multi-regions, different technologies, more general constraints, a stochastic approach and a Mixed Integer Programming (MIP) formulation of learning curves. PSI developed a new improved version of ERIS (Energy Research and Investment Strategy) prototype model, with the two-factor learning curve formula (¿2FLC¿) that includes learning by doing and learning by searching. The model was applied to study the global power generation system. In the endogenous specification of R&D expenditures a global R&D budget is provided and the model finds its optimal allocation among the different learning technologies taking into account a number of intervening factors. In addition, in order to reflect the fact that time lags and depreciation occur in the R&D process, a knowledge stock function was introduced in place of cumulative R&D expenditures.
The systems-engineering model MESSAGE is an energy system cost optimisation model for the world under perfect foresight. The original model was augmented to include induced technological progress. The analysis with MESSAGE covers issues of the very-long-run optimal configuration of the World energy system. The IIASA modelling team developed a set of global energy-environment scenarios of high economic and energy demand growth in which technological change unfolds in alternative "path dependent" directions. The scenarios share common demographic, economic, and energy demand developments, but explore alternative development pathways for the technology mix and future carbon emissions. Enhanced expertise was gained through the analysis and the assessment of long-term global and world-regional E3 (energy-economy-environment) scenarios. Such global scenarios can serve as a background for world-regional or national studies; past examples of such activities and related results include contributions to the IPCC and collaboration with the World Energy Council (WEC). Finally, MESSAGE was used to estimate the impact of R&D expenditures on specific electricity generation technologies for two R&D objectives; (1) the mitigation of cumulative CO2 emissions, (2) the reduction of consumer costs.
Central to the SAPIENT approach is the elaboration of a small aggregated model named ISPA (Integrating System for Priority Assessment), specifically designed for R&D budgeting policy exploration. In essence, the ISPA model explores a domain of optimal R&D strategies in a context of uncertainty (i.e. incorporating notions of hedging) and in the presence of multiple objectives as is appropriate when considering public sector participation in R&D initiatives. In recognition of the nature of such policy, namely that it pursues a number of aims, and that R&D is essentially a speculative activity, ISPA takes the form of a multiple objective stochastic non-linear optimisation problem where the probability that an objective exceeds a given desirable threshold is maximized subject to the condition that the probability that the other objectives exceed given thresholds is greater than a certain level. Naturally the budget allocations must also be kept non-negative. A concrete specification of this problem has been coded and tested here in view of gaining insights for a method of policy option explorations.
The technology Improvement Dynamics Database is a full set of data on 23 power generation technologies following the specification of the POLES model classification and covering world installed capacity, sales in volume and value, estimates of R&D expenditure split into public and private devoted to each technology together with time series on the evolution of their costs. It constitutes at present a unique information set which beyond being utilised as the common data base for SAPIENT project, has considerable potential for interesting international organisations dealing with technology and environmental issues as well as the main European companies in the power generation equipment and electricity sectors.
ECN improved the model representation of induced technological progress in MARKAL by the inclusion of learning in clusters of technology in the ANSWER user-interface of MARKAL. Additional features were introduced to MARKAL in order to obtain the information relevant and requested for the ISPA model (e.g. cumulative sales, CO2 abatement costs). Regarding MARKAL applications ECN has made scoping calculations for solar PV and wind energy to show the feasibility of its indirect two-factor (i-2FLC) learning curve approach for the EU MARKAL model. The R&D information collected by IPTS was extensively utilised in order to fit the relationship between R&D intensity and technology progress ratios. Regarding the centralised policy applications ECN developed and tested a procedure to run MARKAL many times in order to produce multiple run output as input to ISPA. Going beyond proposal requirements but staying at the heart of the SAPIENT project�s main area of concern, ECN combined the latter facility with a Monte Carlo uncertainty analysis focussing on uncertainties in the parameters driving the induced technology progress ('endogenous learning').
ECS-IIASA has developed a number of global energy-economic-environmental (E3) scenarios. These include reference scenarios contributed to the recently published Special Report on Emissions Scenarios (SRES, 2000) of the Intergovernmental Panel on Climate Change (IPCC). In this context ECS-IIASA has undertaken a comparison based on the main scenario indicators for some POLES (REF 1999 and 2001) and MESSAGE (A1B for SRES and, mainly in the summary, A1B-550) scenarios. The analysis has highlighted the main common points and divergence with respect to scenario results on the scenarios� primary energy mix, electricity generation, carbon emissions, as well as carbon and energy intensities putting particular emphasis on the medium term paths in the context of longer term trajectories. Some of the results indicate that the three principal measures contributing to the goal of limiting atmospheric GHG concentrations are: (1) demand reduction due to enhanced energy conservation, (2) fuel switches away from carbon-intensive fuels, and (3) carbon scrubbing and sequestration. The full results of this comparison can be found in the final report.