Because PV arrays are always installed in poor outdoor environment, a variety of faults often occur during the operation. In order to obtain the types of fault, a fault diagnosis method of the BP neural network based on L-M algorithm is proposed. Through the in-depth analysis of the output of the PV array under normal state and fault states, the input variables of the diagnosis model are obtained. Compared with other fault diagnosis methods for the PV array, the proposed method does not need additional equipments, so the cost is reduced and the system can be run online and real-time. Finally, the simulation and experimental results show that the fault diagnosis method for the PV array based on the BP neural network can effectively detect four types of fault for PV array such as short-circuit, open-circuit, abnormal degradation and partial shading.