Research On Fault Diagnosis Method Of Photovoltaic Array Based On Improved Autoencoder

The fault diagnosis of photovoltaic arrays is of great significance to the safe and stable operation of photovoltaic power stations.This thesis takes the output current and output voltage of a photovoltaic array as the starting point of fault diagnosis,studies the output characteristics of the photovoltaic array under different operating conditions,and proposes a photovoltaic array fault diagnosis method based on an improved autoencoder,which realizes the automatic extraction of photovoltaic array fault features and accurate detection of fault types.(1)Analyze the single diode mathematical model of a photovoltaic cell,the smallest component unit of a photovoltaic array,build the simulation model of the photovoltaic module and photovoltaic array,and determine the "four parameters" of the photovoltaic module.Under the condition of considering the nonlinear output of the photovoltaic array,analyze the typical fault types and fault causes of the photovoltaic array,and simulate seven different typical operating conditions.A total of 15 different types of fault samples were collected,and the output characteristics of photovoltaic arrays under various working conditions were analyzed,which provided the data basis for the subsequent fault diagnosis of photovoltaic arrays.(2)Aiming at the problem that the traditional photovoltaic array "four parameters" cannot extract more important fault features,which leads to low fault diagnosis accuracy,we proposed a photovoltaic array fault diagnosis method based on Dropout optimization stack auto-encoder(SAE).Since SAE’s deep stacked network structure superimposes multiple hidden layers,the model parameters increase.To prevent overfitting of the model during training,we introduce the Dropout method and Adam algorithm to optimize the stacked autoencoder.The experimental results show that the optimized fault diagnosis model can not only automatically extract the fault features of photovoltaic arrays,prevent the occurrence of the overfitting phenomenon,but also accelerate the convergence rate of the model and accurately identify the fault types of photovoltaic arrays.(3)Aiming at the problems that the output data of a photovoltaic array has strong timing and the autoencoder model is initialized randomly with super parameters,a fault diagnosis method of photovoltaic array based on Bayesian optimization is proposed.The method introduces the context feature extraction module and the convolutional attention mechanism based on the autoencoder.Thus,the ability of the fault model to automatically extract fault timing features is enhanced,and the Bayesian Optimization Algorithm(BOA)optimizes the model super parameters.The experimental results show that the improved convolutional autoencoder can accurately identify the operating conditions of photovoltaic arrays,and has a higher accuracy than other traditional fault methods.It provides a research scheme for the optimization of the autoencoder model and has good theoretical research significance and practical value.