Coordinated Distributionally Robust Optimal Allocation of Energy Storage System for HV-MV Distribution Network Resilience Enhancement

To solve the problem of power imbalance under extreme and normal scenarios in high voltage (HV) and middle voltage (MV) distribution networks with high penetrations of photovoltaic (PV), the paper proposes a distributionally robust optimal allocation method of energy storage system (ESS) for equilibrating resilience and economy. Firstly, due to the strong relation between resilience enhancement and ESS location, the siting-sizing sequential updating method is adopted to site ESS based on the planning resilience indexes. Secondly, to generate normal and extreme PV-load scenarios with continuous labels, an adversarial autoencoder (AAE) combined with transfer learning is improved by introducing the conditional neural network. Based on the improved AAE and clustering techniques, a Kullback-Leibler divergence-based ambiguity set is constructed to characterize the joint probability distribution of PV and load. Thirdly, considering the boundary information interaction of HV-MV distribution networks, a two-stage coordinated distributionally robust optimization model is established to size ESS. In the model, centralized ESS in HV distribution network participates in frequency regulation and peak shaving while distributed ESS in MV distribution network regulates voltage profiles in the four-quadrant mode. For further improving resilience, an operational resilience index as a chance constraint considering extreme scenarios is embedded into the model. Fourthly, the C&CG algorithm is nested into the analytical target cascading method to handle the model. Finally, case studies show that the proposed method can effectively balance resilience enhancement and economic operation.