Research On Online Fault Detection Of Photovoltaic Array Based On GA-BP Neural Network

Solar energy is the most promising renewable energy,photovoltaic power generation technology has attracted the high attention of countries around the world,China’s photovoltaic industry has ushered in explosive growth since 2014,but photovoltaic power generation system is vulnerable to environmental factors in actual operation and resulting in the failure of photovoltaic arrays.In order to reduce the loss of power generation and improve the reliability of photovoltaic power generation system operation,how to quickly and accurately complete the detection of photovoltaic array failure is the current urgent need to solve the problem,based on this,the research conducted in this paper is as follows.(1)Study the working principle and output characteristics of photovoltaic cells.The working principle is studied by analyzing the mathematical model of PV cell and its equivalent circuit structure.On this basis,the relationship between the output characteristics of the cell monomer and light and temperature during normal operation is investigated,and the changes of typical parameters of the cell are determined.(2)Establish a PV array model to simulate the types of faults that may occur in the operation of the actual array and collect fault data.According to the in-depth study of the power generation principle of PV cells and the connection structure of the array,combined with the fieldwork in Zichuan Xinmingzhu PV power generation center,a simulated PV array with four rows and three columns is established in MATLAB/Simulink,so as to simulate the common fault states of PV arrays under different fault conditions,to study the causes of array faults,to analyze the output characteristics of the array under each fault state compared with the normal fault-free condition,and real-time fault operation data are obtained.(3)Feasibility study of GA-BP neural network in PV array fault detection.This paper uses Genetic Algorithm(GA)to select the initial weight threshold of BP neural network based on the analysis of the structural composition and specific parameter selection of BP neural network,and establishes the fault diagnosis model of PV array based on GA-BP neural network on account of the problem that the random selection of the initial weight threshold of BP neural network interferes with the fault detection results.The model is used to detect four kinds of faults that often occur during the actual operation of PV arrays.(4)In order to further improve the detection performance of GA-BP neural network for photovoltaic array fault types,a hybrid genetic algorithm integrated with Pigeon-Inspired Optimization Algorithm(PIO)was proposed to optimize BP neural network.The different combination methods of the two algorithms are analyzed,conducts in-depth research on the possibility of the integration of the two optimization algorithms,and analyzes the original parameters of the pigeon flock optimization algorithm were further improved.(5)The performance difference between the BP neural network optimized by hybrid genetic algorithm and the original GA-BP neural network in photovoltaic array fault detection is studied.Comparative analysis of the experimental results shows that the number of iterations of the improved GA-BP neural network is much less than that of the original GA-BP neural network,which effectively shortens the detection time and improves the rapidity of the fault detection model,and the improved GA-BP neural network achieves a comprehensive detection accuracy of 98.7%for various fault states,and is also more accurate for the detection of aging faults and shadow obscuration faults.