TSO-DSO Interaction: Privacy-Preserving Optimal Power Flow with Distributed Generators Using a Machine Learning-Based Approach

The integration of distributed generators (DGs) into the electrical grid has introduced new dimensions in power system management: Enhancing efficiency and operational flexibility of the grid requires taking into account the DGs located at the distribution level in dispatch decisions, while also considering the operational constraints of both distribution and transmission levels. However, the centralization of cooperation between transmission system operators (TSOs) and distribution system operators (DSOs) is troublesome due to privacy concerns of both parties. In present paper, we introduce a Machine Learning-Based Privacy-Preserving Optimal Power Flow (MLB-PP-OPF) framework. This method enables the seamless inclusion of DGs at the distribution level in optimal power flow calculations, all while adhering to the operational limits of both systems and maintaining data privacy. The proposed approach leverages three machine learning models to map the distribution system variables to active and reactive power flows at the point of common coupling, and to determine the feasible operating region of the distribution system in a privacy-preserving manner. Comparative case studies, performed on the coupled IEEE 30-bus transmission and 33-bus distribution test systems, demonstrate that the results achieved using MLB-PP-OPF align almost perfectly with those obtained through classical OPF, while maintaining data privacy.