Sustainable power systems operations under renewable energy induced disjunctive uncertainties via machine learning-based robust optimization

Abstract

For sustainable and reliable power systems operations integrating variable renewable energy, it is essential to incorporate the uncertain intermittent power outputs. A novel robust unit commitment framework with data-driven disjunctive uncertainty sets is proposed for sustainable energy systems with volatile renewable wind power, assisted by machine learning. The approach can flexibly identify the uncertainty space based on renewable power forecast error data with disjunctive structures. Specifically, the uncertainty data are grouped using K-means and density-based spatial clustering of applications with noise (DBSCAN) following the optimal cluster number determined by the Calinski-Harabasz index. Disjunctive uncertainty sets are constructed accordingly as the union of multiple basic uncertainty sets, including conventional box and budget uncertainty sets, and data-driven uncertainty sets using Dirichlet process mixture model, principal component analysis coupled with kernel density estimation, and support vector clustering. Subsequently, the problem is formulated into a two-stage adaptive robust unit commitment model with data-driven disjunctive uncertainty sets and with a multi-level optimization structure. To facilitate the solution process, a tailored decomposition-based optimization algorithm is developed. The effectiveness and scalability of the proposed framework are illustrated with two case studies investigating sustainable operations scheduling based on IEEE 39-bus and 118-bus systems, considering the integration of intermittent renewable energy. Results show that the proposed framework can reduce the price of robustness by 8–48% compared to the conventional “one-set-fits-all” robust optimization approaches. Benchmarking with stochastic programming indicates that the proposed framework can achieve the same or better economic performance for sustainable operations with over 75% less computational time.