Data-Driven Distributionally Robust Unit Commitment With Wasserstein Metric: Tractable Formulation and Efficient Solution Method

摘要

In this letter, we propose a tractable formulation and an efficient solution method for the Wasserstein-metric-based distributionally robust unit commitment (DRUC-dW) problem. First, a distance-based data aggregation method is introduced to hedge against the dimensionality issue arising from a huge volume of data. Then, we propose a novel cutting plane algorithm to solve the DRUC-dW problem much more efficiently than state-of-the-art. The novel solution method is termed extremal distribution generation, which is an extension of the column-and-constraint generation method to the distributionally robust cases. The feasibility and cost efficiency of the model, and the efficiency of the solution method are numerically validated.