Strategic investment in electricity markets: Robust optimization versus stochastic programming

Decarbonization policies have spurred the adoption of variable renewable energy (VRE) technologies such as wind and solar power. To enable flexible resources and accommodate VRE’s intermittency, electricity markets are shifting toward renewable-aware dispatch based on stochastic optimization. However, strategic firms may exploit such market structures to manipulate prices to their advantage. To complement the extant literature, we compare investment decisions as well as worst-case profits and losses in the context of generation expansion by a strategic firm that uses either risk-averse stochastic programming or robust optimization. The former is a bi-level optimization problem, whereas the latter is a tri-level problem. Our contributions are threefold in addressing policy and methodological challenges. First, we demonstrate that using robust optimization instead of stochastic programming generally leads to investment plans with a higher share of VRE because it serves as a hedge during undesirable realizations with low consumer willingness to pay and high marginal costs for conventional generation. Second, a regret analysis shows that the worst-case profit is significantly reduced if an investor uses expansion decisions from stochastic programming, highlighting the importance of selecting a methodology aligned with the main objective of the investor. The effect is especially pronounced if decisions stem from a social planner, thereby indicating how a conventional, centralized perspective may fail to reflect private incentives for generation expansion in evolving electricity markets. Third, the analysis of strategic behavior necessitates state-of-the-art decomposition techniques such as the constraint generation-based algorithm and the column-and-constraint generation algorithm for the bi- and tri-level problems, respectively. • Strategic two-stage generation-expansion planning under uncertainty is addressed. • Novel strategic robust optimization with recourse formulation is presented. • Robust optimization and risk-averse stochastic programming frameworks are compared. • Exact solution procedures are applied. • Unsuitable modeling tools lead to significant reductions in the profit.