Abstract: In order to further improve the output efficiency and dynamic control performance of the photovoltaic power generation system,an improved perturbation observation method MPPT control is designed in this paper based on neural network. A neural network algorithm is introduced based on the perturbation and observation method,so the dynamic adjustment speed is accelerated when the environment changed drastically. Through system simulation and hardware measurement,it’s concluded that,compared with the traditional perturbation observation and neural network method,the dynamic adjustment speed can be accelerated and steady error can be reduced to some extent with the improved perturbation observation method.