Recognition of solar cell modules defects based on optimized BP neural network

摘要

Since the standard BP algorithm to recognize the defects of solar cell modules have the problem of slow convergence speed and easily falls into relative minimum.The recognition method which based on optimized algorithm is proposed.After preprocessing image of electroluminescence infrared image of solar cell component in debris,crack,off-grid,open weld and black pieces,extracting 5 combined invariant moments of the defect as the input of BP neural network.Then network of the BP algorithm with additional momentum,L-M algorithm and standard BP algorithm are used to performance comparison in recognition system.The results indicate that,the BP neural network of L-M algorithm have the advantages of fast convergence speed and high accuracy.The recognition rate of 5 kinds of defect has reached more than 90％.