A Neural Network Fault Diagnosis Method For Photovoltaic Power Generation System

Under the influence of complex environment and other factors,photovoltaic power generation system failures frequently occur.Simple monitoring and fault diagnosis technology can not realize the intellectualization and informatization of the system.Photovoltaic power generation system is a complex nonlinear system,which has the characteristics of time-varying and uncertainty.It is difficult to obtain an accurate mathematical model,so it is suitable for artificial intelligence diagnosis.When the photovoltaic system is in a different state of failure,the output characteristics of each component will be different.The neural network provides a method that can accurately map out the mapping relationship between different fault states and the output characteristics of the system.According to the structure of the photovoltaic system,shortcomings and difficulties of current research about fault diagnosis of small independent photovoltaic systems are analyzed and summarized,and we propose a general neural network fault diagnosis method that detects diagnosis signal which reflect running state of photovoltaic power plants,using neural network method for fault type identification.Based on the self-made micro photovoltaic power station,experiments are carried out in the indoor and outdoor environment,and the proposed method is verified.The major research work and innovation point of Master thesis are as follows:1.The types of faults and causes of faults in domestic and foreign literatures are summarized.In combination with the actual situation of photovoltaic power plants,photovoltaic modules with high failure rates are used as typical case studies.As the independent photovoltaic power generation system has energy storage links,it has a compensating effect on the output of the faulty system,and the existing fault diagnosis methods have certain limitations.2.Aiming at the fault diagnosis problem of independent photovoltaic power generation system,the mathematical problem description of neural network fault diagnosis is established.On this basis,a general neural network fault diagnosis strategy for PV system is proposed.BP neural network(BPNN)and Elman neural network(ENN)as two specific models were embedded in neural network fault diagnosis strategy,forming two specific neural network fault diagnosis methods of PV system.At the same time,in the light of the problem of the lack of fusion between the existing neural network method and the fault state of PV system,three optimization strategies,Diagnostic signal type selection strategy,neural network optimization training strategy and initial weight optimization strategy,are added to enhance the performance of fault diagnosis.3.According to the existing literature only by simulation software or single block photovoltaic panel experiments,experimental conditions and environment is simple,lack of abundant fault information input type,so this experiment is conducted in a laboratory and outdoor real environment using a self-made photovoltaic power plant simulation platform.At the same time,multiple fault types are set up,and the existing methods are compared with the effectiveness of the proposed method.