Accommodation Evaluation Method for Renewable Energy Power Systems Based on Lagrange Soft Actor-Critic Deep Reinforcement Learning

Currently, the installed capacity of renewable energy has become the primary source in the power system, while its power generation and energy output are gradually transitioning to the major share. However, due to its inherent randomness and volatility, the renewable energy curtailment has become a critical issue. Therefore, evaluating the accommodation of renewable energy under diverse power output scenarios is necessary. To address this challenge, in this paper, a deep reinforcement learning (DRL) model is firstly designed for renewable energy power systems, which includes the accommodation evaluation model, agent action space, agent state space and reward mechanism. A DRL framework based on the Lagrange Soft Actor-Critic (Lag-SAC) algorithm is then proposed, elaborating on its foundational theoretical model and core network modules. Finally, the solution process for system accommodation evaluation using the Lag-SAC algorithm is introduced. A case study of a wind-photovoltaic-thermal-storage power system is conducted to validate the effectiveness and adaptability of the Lag-SAC algorithm in system accommodation evaluation.