A master-slave residual network-based classification method for PV panels dust accumulation

In this paper, considering the idea of the master-slave structure, a new classification method based on master-slave residual network is proposed to classify the degree of dust accumulation on the photovoltaic (PV) panels. Through training the master-slave residual network, the structure parameters of residual network can be determined, such that the proposed classification method can obtain the better classification accuracy for different learning tasks. In order to determine the structure parameters of residual network, four learning algorithms are given to determine the input type, the number of convolution layers, weight learning method and the pooling method, respectively. In order to verify the accuracy of the classification method, an image dataset for different degree of dust accumulation is collected by artificial simulation experiment, and thus, the simulation results show the effectiveness of this proposed method.