Data-driven affinely adjustable distributionally robust framework for unit commitment based on Wasserstein metric

This study proposes a data-driven distributionally robust framework for unit commitment based on Wasserstein metric considering the wind power generation forecasting errors. The objective of the constructed model is to minimise the expected operating cost, including the generating cost, start-up and shut-down costs, and also the reserve cost, which overcomes the shortcomings of the conventional model without optimising the reserve capacity. What is more important, different from the conventional robust optimisation methods, wind power big data is fully utilised in this model to construct the ambiguity set without any presumption about its probability distribution. This is realised by Wasserstein ball with an empirical distribution as the centre. Thus, the proposed robust model is actually data-driven and can immunise the solutions against the worst-case distribution in the ambiguity set. In addition, the scale of the historical data is very critical for this method, the larger the scale is, the smaller the ambiguity set is and the less conservative the result is. Numerical results and Monte Carlo simulations on a real 75-bus system demonstrate the superiority of the proposed model.