Fault Diagnosis Method of Smart Substation Secondary System Based on Deep Reinforcement Learning

Abstract

In view of the problem that most substation faults can only be handled by artificial experience, a fault diagnosis method of secondary system in smart substation based on deep reinforcement learning is proposed. Firstly, the principal component analysis method is used to reduce the data dimension of the secondary system signal in order to improve the data processing efficiency. Then, the parameters of recurrent neural network (RNN) are optimized by deep Q network (DQN), and the RNN-DQN fault diagnosis model is constructed. Finally, RNN-DQN model is used to process the historical state data of long-term operation of smart substation to obtain the fault diagnosis results. Taking a typical 220kV smart substation as an example, the experimental results show that the diagnosis accuracy of RNN-DQN can reach 94%, and the curve fitting effect is good. At the same time, the average error and maximum error of fault diagnosis of the proposed method are 0.008 and 0.601 respectively, which can meet the actual field requirements.