Research on Optimization Model of Machine Learning in Informationized Operation and Maintenance of Power System

This paper presents an optimized model for the management and maintenance of power system based on DRL and GA. The model realizes the real time decision in the large electric power system by means of the deep reinforcement learning technique, and the optimization method is used to solve the problem. This paper presents several test cases, and verifies the validity of the model based on the actual running and maintenance data of a power system. The experiment results indicate that the optimized model has remarkable advantages: firstly, the running and maintenance costs are reduced by 15.3% compared with the conventional optimization method; the failure rate is reduced by 20.7%, the fault prediction precision is increased 12%, and the system stability is improved by 17%, and the repair time is reduced 30%. It is proved that the combination of SLI and GA cannot only improve the running and maintaining efficiency, but also reduce the cost of running and maintaining. Moreover, the adaptive capability of the model makes it possible to react flexibly to the changes of the power system in different working conditions.