Joint Optimization Dispatching for Hybrid Power System Based on Deep Reinforcement Learning

With large scale renewable power integrating, hybrid power system needs joint optimization dispatching. Considering complementary characteristics of different power types, this paper first constructs a day ahead time scale optimized dispatching model. The objectives are minimizing the system operation cost and maximizing the renewable energy consumption. The startup-stop status of thermal units and power output of different type power stations are selected as optimization variables. The problem then is modeled as a multistep Markov decision process which is a sequential decision process problem. A reinforcement learning method, Deep Deterministic Policy Gradient algorithm, is introduced to solve the decision problem. Finally, simulations have been carried out to validate the effectiveness of the proposed method. Numerical results show that the proposed method obtains a satisfied result which can meet the power load demanded, ensure the consumption of renewable energy and minimize the system cost meanwhile.