Research On Photovoltaic Power Prediction Based On Optimized BP Neural Network

With the increasing demand for energy in all countries in the world,the problem of environmental damage and resource crisis is urgent,and the solar photovoltaic power generation has the advantages of renewable,pollution-free and not subject to regional restrictions,and all countries have been vigorously developing photovoltaic energy.However,the photovoltaic output power has randomness and volatility and other problems.The power grid is a real-time balance system.The power generation characteristics of the solar photovoltaic power generation system will cause damage to the power grid when it is connected to the grid,affecting the normal operation of the power grid.Accurate PV output power prediction can reduce the impact on the power grid,so the research on PV output power prediction has very important practical significance and application value.This thesis first elaborates the working principle of the power of photovoltaic cells,photovoltaic power generation system are introduced in the basic composition,working principle,and then in-depth analysis of the output characteristics of photovoltaic power system,selection of the intensity of solar radiation,ambient temperature,wind speed,humidity,and sunny,cloudy and rainy day three weather types as the main factors influencing the photovoltaic power generation system.Based on the analysis of the factors affecting the power generation of photovoltaic power generation system,BP neural network is used to predict the short-term power of photovoltaic power generation.Firstly,the network structure is designed according to the difficulty of the task and the size of the training data.Due to the intensity of solar radiation and photovoltaic short-term power actual values are very obvious,a BP neural network based photovoltaic prediction model is designed,in which the input is influential factors of photovoltaic short-term power,and the output is the short-term power prediction.However,due to the nature of the BP neural network,the prediction accuracy is not high.Aiming at the problem of BP algorithm,genetic algorithm(GA)was used to optimize the initial weight and threshold of BP neural network,and GA-BP photovoltaic short-term power prediction model was established,so that the efficiency and accuracy of power prediction were improved.Then,based on the existing BP neural network-based short-term photovoltaic power prediction method,a BP neural network model(BAS-BP)based on the optimization of longicorn beetles search algorithm was established and applied to short-term photovoltaic power prediction.Finally,this paper through the experiment of BP and GA-BP and BAS-BP three photovoltaic power prediction model contrast test,the experimental results show that the BAS-BP prediction model in sunny day,cloudy and rainy days under three types of weather prediction effect is superior to the traditional BP neural network forecasting model and GA-BP prediction model,even under complex weather conditions,the prediction model based on BAS-BP remained higher prediction precision,effectively improve the accuracy of the short-term photovoltaic(pv)power output prediction,and to overcome the long training time,the shortcomings of slow convergence speed.