Research On Fault Detection Of Photovoltaic Array Based On Image Processing

By intelligently processing and analyzing the collected photovoltaic array images through image processing technology,the fault detection of photovoltaic array can be realized in a non-contact,high convenient and high precision way,which is of great significance for maintaining the safe and efficient operation of photovoltaic power plants.In this paper,a fault detection method combining visible light and thermal imaging images is adopted.On the basis of detecting internal faults,the method can eliminate interference such as sunlight reflection and shielding.Specific work contents include:(1)With the aim to solve the problems of noise interference and geometric distortion in the collected photovoltaic array images,bilateral filtering and perspective correction are respectively used.An improved two-dimensional maximum betweenclass variance method is proposed,which used the edge information of the filtered image to divide the histogram area,and then the image is segmented by the intercept threshold method,and good results are achieved.Then through the separation of the target area,the interference of ground background and other factors are reduced.(2)Analyzed the image characteristics of the photovoltaic array under several abnormal conditions,and introduced the commonly used image features and feature extraction methods.(3)In order to improve the accuracy of the classification model based on the support vector machine,the improved state transition algorithm is used to optimize its parameters.To solve the problem of uneven and unstable initial population distribution of the state transition algorithm,logistic mapping is introduced.In addition,with the aim to improve the algorithm global search and convergence capabilities,two improved strategies are proposed.One is to introduce the levy flight mechanism into the random step size,the other is the normal distribution with adaptive variance.(4)Introduced the construction method of the photovoltaic array,the selection of image acquisition equipment and shooting method.K-means algorithm is used to distinguish the fault,the board surface and other areas.The average value of the board boundary is used to divide the battery grid,and the number of the faulty battery is obtained through position matching.The experimental results showed that the photovoltaic array fault detection method based on image processing adopted in this paper has good feasibility and effectiveness,and it is beneficial to improve the operating efficiency of photovoltaic power plants and reduce maintenance costs.