Learning Active Constraints to Efficiently Solve Linear Bilevel Problems: Application to the Generator Strategic Bidding Problem

Bilevel programming can be used to formulate many problems in the field of power systems, such as strategic bidding. However, common reformulations bilevel mixed-integer linear programs make solving hard, which impedes their implementation real-life. In this paper, we significantly improve solution speed and tractability by introducing decision trees learn active constraints lower-level problem, while avoiding introduce binaries big-M constants. The application machine learning reduces online time, moving selection an offline process, becomes particularly beneficial when same problem has solved multiple times. We apply our approach bidding generators electricity markets, where solve times for varying load demand or renewable production. Three methods are developed applied a generator, with DCOPF lower-level. These heuristic so, do not provide guarantees optimality quality. Yet, show that networks sizes, computational burden is reduced, also manage find solutions were previously intractable.