Research On Fault Diagnosis Of Photovoltaic Array Based On Machine Learning Algorithm

In view of the increasingly prominent problems of energy security and global climate,the development and utilization of renewable energy is imperative.Solar energy has been widely concerned around the world because of its large reserves,green and pollution-free advantages.Photovoltaic power generation is the main way of solar energy development and utilization.With the continuous development of photovoltaic power generation technology,the number of photovoltaic power stations in the world is increasing.Due to the complex working environment of large-scale photovoltaic array and the large number of panels,it is inevitable that there will be various failures in the operation process,which seriously affects the work efficiency of the whole system.Therefore,the real-time effective diagnosis of photovoltaic array fault is the premise to ensure the efficient operation of photovoltaic power generation system.In this paper,photovoltaic array is taken as the research object,through the analysis and research of four typical faults,namely short circuit,open circuit,aging and local shading,the researches in photovoltaic array fault diagnosis are carried out based on machine learning technology.The main contents of this paper are as follows:(1)Matlab/Simulink software is used to build the photovoltaic array fault simulation model.Firstly,based on the brief analysis of the working principle and equivalent circuit of photovoltaic cells,two modeling methods are used to build the simulation model of photovoltaic modules,and the better module model is selected to build the photovoltaic array fault simulation model.Then,the local shading,open circuit,short circuit and aging state of photovoltaic array are simulated by changing the light intensity and resistance value.Finally,the output parameters of photovoltaic array under typical fault conditions are analyzed in detail,and the input characteristic parameters of the fault diagnosis model are determined.(2)Two common machine learning algorithms in the field of artificial intelligence are selected: least squares support vector machine(LSSVM)and kernel extreme learning machine(KELM).Firstly,in order to improve the fault diagnosis accuracy of LSSVM,the whale optimization algorithm is used to optimize the parameters of LSSVM,and a fault diagnosis method based on WOA-LSSVM is proposed.Then,aiming at the problems that bat algorithm is easy to fall into local optimum and slow convergence speed in the later stage,tent mapping and gaussian perturbation strategy are introduced to improve bat algorithm,and a TGBA algorithm is proposed.The regularization coefficient and kernel parameters of KELM are optimized by TGBA algorithm,and the best combination of parameters is obtained.The fault diagnosis model based on TGBA-KELM is established.(3)The hardware experimental platform of photovoltaic array is built,and the experimental simulation of short circuit,open circuit,aging and partial shading fault is carried out on the experimental platform.The experimental data of normal operation and faults of photovoltaic array under different weather conditions is collected.Secondly,the accuracy and effectiveness of WOA-LSSVM method and TGBA-KELM method are verified by using the collected experimental data.Finally,the simulation test is carried out by using WOA-LSSVM,TGBA-KELM and BP algorithm.The experimental results show that TGBA-KELM has higher diagnosis accuracy and better stability.