Instance generation and Benders' decomposition for long-term expansion planning problems

The expansion planning (EP) problem in energy systems optimization aims to determine long-term investment decisions while ensuring the feasibility of short-term system operations. Long-term (investment) decisions include capacity expansion, asset decommissioning, and lifetime extension, whereas short-term (operational) decisions concern generation scheduling, which must satisfy the supply–demand balance subject to technical and network constraints.

This presentation addresses EPs from two complementary angles. First, after a "brief" review of the literature on EP and an introduction of our case of study, we present the European Resource Adequacy Assessment (ERAA) dataset, which, combined with complementary sources such as RTE reports (e.g, Futurs énergétiques), provides a basis for generating realistic EP instances. Second, we focus on the solution method, namely Benders decomposition, which is widely used to solve such problems. In particular, we examine how to select a cut, among a set of candidates, to separate a given master problem's solution. We review several cut selection policies from the literature, along with the properties a cut may satisfy, and discuss how these properties relate to convergence and computational performance. We conclude with a numerical analysis based on instances generated from the ERAA dataset, assessing the impact of different cut selection strategies on the performance of the Benders decomposition algorithm.