Research On Fault Diagnosis Method Of Photovoltaic Modules Based On Deep Learning

The rapid development of the photovoltaic power generation industry is of great significance to the realization of the dual carbon goal.For the safe operation of the photovoltaic power generation system,the establishment of a complete,reliable and rapid detection system is the key to ensure the safe operation of the photovoltaic system.The most important part of the photovoltaic system is the photovoltaic module;which health will affect the operating state of the system.So the fault diagnosis of the photovoltaic module is particularly important.The amount of operating data of photovoltaic modules is large,and deep learning has the ability to process large amounts of data.However,the establishment of deep learning frameworks is highly dependent on experts’ experience,and its design is time-consuming,which can no longer meet the requirements of photovoltaic fault diagnosis under mass of data.Therefore,this thesis proposes an intelligent diagnosis method based on convolutional neural network combined with improved gate control unit,which can complete fault feature extraction and fault identification and diagnosis automatically.(1)Firstly,the mathematical model of photovoltaic cells is established.Then,the photovoltaic module faults are analyzed,and simulations are performed for open circuit,short circuit,aging,bypass diode failure,and different shadow failures.Finally,a 4×4 photovoltaic array is built,and the output characteristics of the array is simulated and explained.(2)A one-dimensional convolutional neural network framework is proposed according to the characteristics of the data by analyzing the operating data of photovoltaic modules,which introduce an adaptive batch normalization method for the optimization model and improve 1DCNN.According to the timing characteristics of the running data,a Gated Recurrent Unit is introduced and improved,and its sensitivity to timing information is used to improve the model’s ability to process data.Therefore,a fault diagnosis method for photovoltaic modules based on1DCNN&bidirectional GRU is proposed.The experimental results show that this method is more effective in classification and anti-noise.(3)Aiming at the small sample problem in the process of photovoltaic module diagnosis,the data generation capability of the Generative Adversarial Network is used to enhance the data.By improving the data augmentation framework and introducing semi-supervised learning,unlabeled samples can be utilized well and the dependence on labeled samples is reduced.A method combining generative adversarial networks and semi-supervised learning is proposed for data augmentation.The1DCNN-bidirectional GRU method is used in classification.The final test results show that this method improves the accuracy of the model for fault classification.