Quick Evaluation of Renewable Energy Accommodation Based on Lagrange-Soft Actor Critic Method

With the amounts renewable energy connected with the power system, the evaluation of renewable energy accommodation is important for reducing the redundant capacity of renewable energy unit and improving the stability of power system. This paper proposes a safe deep reinforcement learning-based approach for assessing regional yearly renewable energy accommodation. The proposed method trains an agent to offer the unit commitment of unit cluster for replacing the solving of time-series production simulation. Through the training of reinforcement learning, the agent offline generates available strategy and achieves the quick unit commitment by the mapping of agent. Meanwhile, in the training process, the agent efficiently explores strategy by amounts of scenes with different renewable energy source samples to improve the perception of agent for the uncertainty of renewable energy. For showing the effective and convenience of proposed method in evaluating the renewable energy accommodation, the samples of certain province from State Grid Corporation of China are used to prove it.