Study On Photovoltaic Power Short-term Forecast Based On Improved GRNN

Along with the high-speed development of the world's energy economy,people for the application of traditional energy demand grows day by day,energy in great demand on the one hand,to bring the exhaustion of the non-renewable energy,on the other hand will cause pollution of the environment,destroy the living and working environment.In order to alleviate the environmental pollution problems and the increasingly shortage of the traditional fossil energy,people try to respond positively to find renewable energy and effort.Solar energy is a kind of renewable energy,which has a wide range of distribution,high energy efficiency,easy to be mined and pollution-free,and has obvious advantages over other energy sources.But affected by meteorological factors such as photovoltaic power,has obvious randomness,so you need to improve the prediction accuracy of the photovoltaic power output,the research emphasis of this paper is mainly addresses the relevant forecast method of solar photovoltaic power.First of all,from the photovoltaic power prediction method starting with the necessity and importance of this paper briefly introduces the present world the development of solar photovoltaic industry,the current photovoltaic power prediction algorithm for classification.Principle and system structure of power system are analyzed,the array of photovoltaic power generation system model has carried on the strict mathematical analysis,and on the basis of theoretical analysis,several influence factors on photovoltaic power are described,laid a theoretical foundation for the prediction power.Secondly,this paper makes a comparative analysis of several major solar pv prediction methods and puts forward the evaluation indexes of the prediction methods.This paper analyzes the principle of abnormal data,presents new abnormal data detection method based on time series,and performs abnormal detection of raw data of experiment.Finally,the generalized regression neural network algorithm is studied,because the traditional GRNN has two problems: first,the algorithm runs relatively slowly when the data is large.Second,a single parameter makes it impossible to update the system dynamically.Therefore,this article put forward based on fuzzy c-means clustering algorithm and chaos optimization algorithm,the improved generalized regression neural network forecasting model,to test samples,through this model to verify the rationality of the proposed algorithm in this paper.