Two-level Planning for Novel Power Systems Based on Copula Theory and Deep Reinforcement Learning

This paper proposes a novel power system planning method based on Copula theory and deep reinforcement learning. Firstly, Copula theory is used to solve the correlation and uncertainty of wind and solar power generation. Then, by generating realistic scenarios of wind and solar output, we develop a two-level optimization model that combines investment decision-making and operational simulation for a new power system. The lower-level model is formulated as a mixed-integer second-order cone programming problem, which can be solved using the CPLEX solver. The upper-level model is a decision-making model that employs a deep reinforcement learning algorithm based on the deep deterministic policy gradient to enable dynamic planning decisions in the new power system. Finally, case studies demonstrate the effectiveness of the proposed approach in significantly reducing investment and construction costs for power systems.