Research On Photovoltaic Array Fault Diagnosis And Location Based On Deep Learning And Image Processing

Photovoltaic power generation,as an efficient and clean energy,has increasingly become one of the main sources of electricity.In photovoltaic power plants,photovoltaic arrays are generally exposed to various types of failures in complex,harsh,and volatile environments.In order to ensure stable and efficient operation of photovoltaic arrays,it is necessary to quickly detect the type and location of failures in the photovoltaic array when they occur.In view of the above problems,this paper takes photovoltaic array as the research object,and investigates the problems of PV array infrared image processing in the process of data imbalance and fault location of PV array fault diagnosis:Firstly,build a fault simulation model by analyzing PV arrays that fail in actual operation,which relies on the classification ability of deep Extreme Learning machine(DLEM).The deep limit learning machine algorithm is used as the base model for fault diagnosis.Meanwhile,to solve the problem that the classification accuracy deviates when the key parameters are artificially set in the classification of the deep limit learning machine,the improved sparrow search algorithm is introduced into the deep limit learning fault diagnosis algorithm to find the global optimal parameters to solve the problem of insufficient accuracy of the key parameters.The data obtained from the simulation simulations are processed in combination with the SMOTE algorithm,and the improved deep extreme learning machine classification algorithm is trained using the processed data set to establish the SFSSA-SMOTE-DLEM fault diagnosis model to realize the fault detection of PV arrays.At the same time,comparative experiments are conducted with neural network classification algorithm(BP),random forest method(RF)classification algorithm,and DLEM classification algorithm.The experimental results show that the classification recognition rates are 96.1%,75.1%,79.6%,and 82.1%,respectively.After the photovoltaic array fault diagnosis and classification,in order to ensure timely troubleshooting,it is necessary to determine the fault location and take aerial infrared images of the photovoltaic array to further locate its faults.The most important step in the localization process is the processing of photovoltaic array infrared image.Therefore,in order to solve the problem of poor segmentation and fault edge extraction during infrared image processing,this paper proposes an intelligent processing method for photovoltaic array infrared images based on an improved sparrow search algorithm-Tsallis relative entropy iteration method-an improved Canny image algorithm to improve image processing efficiency and detect fault locations.Through the simulation of the photovoltaic array construction model,the simulation of the actual operating state of the actual photovoltaic array,the construction of the fault diagnosis model and the comparative experimental analysis with various photovoltaic array diagnosis algorithms,experimental verification shows that the fault diagnosis and location model proposed in this paper has high usability for different scale photovoltaic power stations,especially for larger PV plants with higher economy.The detection accuracy of the fault diagnosis model based on SFSSA-SMOTE-DLEM is as high as 96.1%.The research on photovoltaic array fault diagnosis and location based on deep learning and image processing has solved the problems of data imbalance in fault data,poor image segmentation and fault feature extraction in fault location.