Study On Cleaning Planning Of Photovoltaic Power Station Based On State Identification

As the most important way to use solar energy,photovoltaic power generation in the world have been widely used.Xinjiang region as an important new energy producing areas,all sizes of photovoltaic power plants all over the territory of Xinjiang.How to improve the power generation efficiency of these power plants is an important research content to make full use of solar energy.In the process on research of a number photovoltaic power plants in the Xinjiang region found that most of the photovoltaic power plant does not focus on the PV array surface cleaning and make a reasonable plan for this important operation maintenance work,the majority of photovoltaic power plants for the clear once a month.In the course of the research,the on-site staff indicated that the main reason why it was impossible to carry out reasonable cleaning plan was that the attenuation of the output of the PV array could not be accurately judged by the surface fouling or local shading.At the same time on the process of cleaning,the PV array output is still show attenuation,the environment,the weather conditions which are roughly the same two consecutive days the photovoltaic array output there is a big difference.In order to solve the above problems,this paper takes the SVM classification algorithm as the method to realize the surface dust and local shading state recognition of PV array,and study the above problems.A 10 * 10 PV array simulation model with the same size and uniform of the PV array in the sub-region of the photovoltaic power plant was constructed.The existing problems of the current cleaning mode of the photovoltaic power plant were analyzed,and the cleaning method was replanned.The output characteristics of the PV array under local shading are analyzed,and the training set and test set are constructed for the SVM classification algorithm based on the research results and field acquisition data.By contrast,the radial basis function and the kernel function are used as the algorithm kernel function,cross validation and network search as the algorithm parameter optimization method.Through the training set and test set,completed the photovoltaic array surface fouling and local shading state recognition.Based on the results of state recognition,by comparing the fouling state of the PV array surface and the output difference of the PV array under ideal state,the cleaning node of the photovoltaic power plant is determined,and give suggestions on the plan of cleaning of photovoltaic power plant.