Endogenous Technological Change in Modeling the Global Energy System

Technology learning curves: Cost improvements per unit installed capacity in U.S. dollars (1990) per kW versus cumulative installed capacity in MW for photovoltaics, wind, and gas turbines. Sources adapted from MacGregor et al., 1991; Christiansson, 1995; Grubler, 1998.

Research Objectives

The aim of this project is to create a new version of the global energy system model MESSAGE that endogenizes the introduction of advanced energy technologies.

Investment costs of energy technologies decrease as their installed capacities increase, but the relationship is nonconvex and therefore difficult to include endogenously in an energy model. Our work represents a first attempt to combine these nonconvex technological learning effects with a full-sized, global, bottom-up energy-systems model that includes detailed regional resolution and technological representation. The model comprises 11 world regions grouped according to levels of economic development and geographic location. Over 100 technologies from extraction to end-use are also included. The advanced model developed in this project will have the capability to analyze the diffusion of individual technologies, as well as the formation of interactive technology clusters.

Computational Approach

From a mathematical perspective, the problem is one of a large nonconvex stochastic global optimization. It has a well-specified structure, making a solution feasible. Our algorithm combines a global adaptive search algorithm with a simultaneous stochastic drawing approach that makes it possible to approximate the original problem by sequences of linear optimization problems. Uncertainties play a very important role in the proposed approach and are treated by using a nonsymmetric utility (risk) function. The global optimization algorithm running in parallel was developed specifically for this project.

Accomplishments

During the first stage of the project, a global one-region world model was implemented and tested in the NERSC T3E environment.

This model incorporates a technological learning mechanism (increasing returns) for selected technological clusters (mainly hydrogen-based technologies, synthetic fuels, wind, and solar). An extensive set of computer runs was performed in order to identify potential interrelations between technological clusters and their sensitivities to model assumptions, and to develop alternative scenarios of new research, development, and demonstration (RD&D) investments.

Significance

This experiment (using NERSC computer facilities) represents a unique opportunity to achieve a significant breakthrough in modeling and evaluating alternative strategic long-term energy policies affecting global and regional development and the environment. Results of the proposed experiment have the potential to significantly impact recommendations for national RD&D policies, early investment decisions in new technologies, and, ultimately, the future condition of the global environment.

Publications

A. Gruebler and A. Gritsevskii, "A model of endogenous technological change through uncertain returns on learning (RD&D and investments)," presented at the International Workshop on Induced Technological Change and the Environment, 26-27 June, International Institute for Applied Systems Analysis, Austria; The Economical Journal (submitted, 1998).

S. Messner, A. Golodnikov and A. Gritsevskii, "A stochastic version of the dynamic linear programming model MESSAGE II," Energy 21, 9, 775-784 (1996).

N. Nakicenovic, "Technological change and learning," Perspectives in Energy 4, 173-189 (1997).

http://www.iiasa.ac.at/Research/ECS/