Acceleration of AC-Optimal Power Flow Based on Reinforcement Learning for Power System Planning

Planning and management of transmission systems will increasingly need to consider long-term scenarios. However, this requires handling a large number of simulations, and this can be computationally expensive. Standard AC-optimal power flow (OPF) techniques used for these simulations are often quite burdensome in terms of computation. This is particularly relevant when dealing with non-convex cost functions. Our goal is to employ machine learning (ML) techniques to accelerate simulations for large scenarios. The key novelty lies in the formulation of the AC-OPF problem as a Markov decision process (MDP) considering both convex and non-convex cost functions and including the node voltage fluctuation control problem. The IEEE 9-bus and 118-bus systems were used as a basis for simulations in different case studies based on real-world data to demonstrate the method's effectiveness over other traditional approaches. The results obtained are very promising and pave the way for new analysis methodologies.