Research On Fault Detection Method Of Photovoltaic Array Hot Spot Based On Convolutional Neural Network

With the increasingly prominent global energy crisis,the environmental pollution caused by the application of traditional fossil fuels is increasingly serious.Countries around the world pay more and more attention to the development and research of clean energy and renewable energy.It has become a very important link in the sustainable development strategy of our country to study the method of efficient use of new energy and develop pollution-free and pollution-free energy system.Photovoltaic power generation technology in the field of new energy has developed rapidly.But in the process of solar power generation,the hot spot fault of photovoltaic array will seriously affect the normal operation of photovoltaic power generation system and reduce the service life of photovoltaic power generation system.Therefore,in this paper,aiming at the photovoltaic hot spot fault during the operation of photovoltaic array,a method of photovoltaic array hot spot fault detection based on convolution neural network is proposed,which uses convolution neural network to identify the infrared image of photovoltaic array to detect the hot spot fault of photovoltaic array.This paper mainly completes the following contents:(1)In this paper,the development status of photovoltaic power generation industry is described,the formation principle of photovoltaic hot spot fault is analyzed,the existing detection methods of photovoltaic hot spot fault are explored and analyzed,and the detection method of photovoltaic array hot spot fault based on convolution neural network is proposed(2)Aiming at the real infrared image of photovoltaic array captured from the scene,this paper proposes a method of image preprocessing,which includes Canny edge detection,Hough line detection,edge information extraction of battery board and perspective transformation.The original infrared image of photovoltaic array is transformed into a picture set which can be used for convolution neural network training and recognition,and the process of making the picture set into a data set is introduced.(3)According to the characteristics of the infrared image data set of photovoltaic array,a convolution neural network model is built.The convolution neural network model is used to train,study and test the data set,and then the training,study and test results are analyzed,so as to optimize the convolution neural network model to obtain better fault detection effect.After getting the test results,the visualization of the test results is realized,which is convenient for the maintenance personnel to locate the photovoltaic cells with hot spot fault.