Optimal capacity expansion requires complex decision-making, often influenced by technology learning, which represents the reduction in expansion cost due to factors such as cumulative installed capacity. However, having perfect foresight over the technology cost reduction is highly unlikely. In this work, we develop a multistage stochastic programming framework to model capacity planning problems with endogenous uncertainty in technology learning. To assess the benefit of the proposed framework over deterministic optimization, we apply a shrinking-horizon approach to compute the value of stochastic solution. Further, a decomposition scheme based on column generation is developed to solve large instances. Results from our computational experiments indicate substantial potential cost savings and the effectiveness of the proposed decomposition algorithm in solving instances with large numbers of scenarios. Lastly, a power capacity planning case study is presented, highlighting the stochastic optimization's ability to anticipate significantly different expansion and production decisions in low- and high-learning scenarios.
Bibliographical noteFunding Information:
The authors gratefully acknowledge the financial support from the National Science Foundation under Grant No. 2030296 and the Minnesota Supercomputing Institute (MSI) at the University of Minnesota for providing resources that contributed to the research results reported in this paper.
© 2022 Elsevier Ltd
- Column generation
- Endogenous uncertainty
- Multistage stochastic programming
- Technology learning
- Value of stochastic solution