As energy shortages and environmental degradation become increasingly prominent,solar power generation technology has gained widespread attention and application.As one of green and clean energy,solar energy is more and more widely used in daily life.With the widespread application and research of photovoltaic power generation technology,the monitoring and fault diagnosis of photovoltaic power generation system and photovoltaic array working conditions have become more prominent.Currently,there are three major methods for diagnosing faults in photovoltaic arrays: sensor-based,photovoltaic model-based,and artificial intelligence-based methods.The main research contents of the thesis are as follows:1.Put forward two machine learning algorithms that are widely used in the field of artificial intelligence: neural network algorithms and support vector machine algorithms.The two algorithms are analyzed and compared in different aspects.Through comparison,it is found that the use of SVM algorithm for photovoltaic array fault diagnosis has many advantages over neural networks over the use of.2.Analyze the importance of data preprocessing in the process of intelligent fault diagnosis algorithms.Data preprocessing is a process of denoising,repairing and improving sample data sets.Reasonable data preprocessing can continuously improve the model's learning efficiency and its accuracy.By analyzing the photovoltaic fault data,it is determined that the data preprocessing method adopted in this paper is mainly data normalization and hot card filling.3.A comprehensive analysis of the types of faults that occur in photovoltaic arrays depicts that the aging,cracks,shadows,hot spots,open and short circuit faults affect internal mechanism of photovoltaic array power generation.Based on this,the characteristic parameters of faults occurring in the photovoltaic array,and the characteristic parameters of the input model are finally determined as short circuit current,open circuit voltage,maximum power point current and maximum power point voltage.4.A research on the optimization and adjustment of the kernel function,penalty factor and kernel parameters of the support vector machine algorithm has been done.With a combination of the network search method and the K-fold cross-checking method,the adjustment and optimization has been done to the kernel function and parameters of the support vector machine model. |