Design Of Photovoltaic Power Prediction System Based On GA-WNN And GRNN Combined Algorithm

Photovoltaic power generation is greatly affected by weather conditions,which causes the output power of photovoltaic stations to be extremely unstable.When large-scale photovoltaic stations are connected to the grid,it will have a huge impact on the grid,and then affect the stability of the grid,and even cause the loss of people's lives and national economy.In order to reduce the impact and ensure the safe and economic operation of the power grid,it is very important to accurately predict the output power of photovoltaic stations.For this reason,the national energy administration has carried out relevant assessment on the prediction accuracy of photovoltaic grid connected stations in various regions.In order to further ensure the reasonable distribution of resources,the operation of power grid can be safer and more economical,and the assessment of the accuracy of Photovoltaic Station prediction has been strengthened,so it is of great significance to study a prediction system that can meet the assessment requirements of Northwest China.Firstly,considering the importance of data to a prediction model,the historical data of photovoltaic stations are preprocessed.Identify and delete duplicate data in historical data;identify,fill and delete abnormal data;Identify,fill and delete abnormal data;Power curtailment data is identified and deleted according to power curtailment formula,and supplemented and modified based on grey correlation analysis.On this basis,the characteristic engineering of the data is carried out,and some factors related to the output power are selected from the data preprocessed.Then,considering that these factors not only have coupling relationship with each other,but also affect the convergence and efficiency of the model when they are all the inputs of the prediction model,the main components of these factors are carried out Through the analysis,the new influencing factors related to the output power that can represent the original data information are selected to ensure the reliability of the input data.Secondly,combined with the seasonal characteristics of Northwest China,according to the different seasons and weather characteristics,the data can be divided into three weather conditions: sunny day,cloudy day,rainy day and snowy day.In consideration of the strong local processing ability and adaptive learning ability of wavelet neural network,a prediction model based on wavelet neural network is proposed.However,the algorithm is easy to fall into local minimum value then,combining with the strong global optimization ability of genetic algorithm,a prediction algorithm of optimizing wavelet neural network based on genetic algorithm is proposed.Compared with the prediction of wavelet neural network,the prediction accuracy of the optimized algorithm is higher.Then,in order to further improve the prediction ability of the output power of the station and reduce the influence on the accuracy of the prediction model of wavelet neural network optimized by genetic algorithm under cloudy and snowy weather conditions,a prediction method based on generalized regression neural network is proposed.The advantage of this method is that it has a strong prediction ability for the data with strong volatility and small amount of data.Based on these two methods,an algorithm of real-time weighted combination of wavelet neural network and generalized regression neural network based on genetic algorithm optimization is proposed.The algorithm takes the minimum square of real-time prediction error as the objective function,optimizes the real-time weight coefficient and constraint conditions,so as to determine the value of real-time weight.Then the combined algorithm model and the first two models are used Compared with the other two methods,this method has better prediction performance.Finally,based on the combination prediction algorithm,the prediction system is packaged.The system adopts B/S structure,spring + struts + Hibernate(SSH)framework,HTML5 + CSS3 technology in the foreground display,and combined with other functions,its physical framework,system framework and its internal logic framework,a photovoltaic power prediction system is designed to provide phase for the dispatching department and photovoltaic stations Technical support.