Research Of Prediction Of Photovoltaic Power Based On Optimized BP Neural Network

In the rapid development of economy and technology,the energy crisis and environmental pollution are getting worse day by day.Therefore,our country has continuously developed renewable new energy in recent years,solar energy has become one of the major green energy sources in our country,and the use of solar energy for photovoltaic power generation has become the main way of solar energy utilization.However,photovoltaic power generation has the characteristics of randomness and low callability due to solar radiation intensity,atmospheric temperature,cloud cover and other meteorological factors,which makes the photovoltaic power generation forecast more important for the grid control strategy,the grid power dispatching and the improvement of the grid power quality.Firstly,This thesis gives a detailed overview of the concept of photovoltaic power forecasting,the principle of photovoltaic cell power generation,the composition of grid-connected photovoltaic power generation system,and the main factors that affect the speed of light-power generation.Secondly,this thesis focuses on the research of BP neural network(BPNN)in artificial neural network,including the design of hidden layer number and number of neurons in the input layer,output layer and hidden layer and the choice of transfer function.Then,aiming at the shortcomings of BPNN method,such as slow convergence speed,trapped in a local minimum point easily,poor robustness,and so on,several main swarm intelligence algorithms,like genetic algorithm(GA),particle swarm optimization(PSO),and wolf pack algorithm(WPA)are proposed to optimize the weights and thresholds of BP neural network.The wolf pack algorithm is a new swarm intelligence algorithm in recent years,there are many advantages in the WPA,such as the fast convergence,robustness and so on.This thesis especially introduces the working principle of the wolf pack algorithm.And the WPA algorithm will be improved by GA algorithm so as to improve the convergence performance of the WPA.The improved BP neural network is used to optimize the structure and parameters of the neural network.This method can avoid the blindness of selecting the network parameters in order to achieve the efficiency and accuracy of short-term photovoltaic power forecasting.Finally,in order to verify the performance of the GWPA-BP photovoltaic power forecasting model established by the BPNN optimized by improved wolf pack algorithm,not only compares it with WPA-BP and classical BP neural networkphotovoltaic power forecasting model in the prediction results and the relative error,but also with other photovoltaic power forecasting models,like PSO-BP and GA-BP photovoltaic power forecasting model in terms of performance.They are respectively established by other swarm intelligence algorithms,such as particle swarm optimization algorithm and genetic algorithm for BPNN optimization.The photovoltaic power data provided by the UQ Solar project contest is used as the experiment sample to carry out the simulation experiment,and predict the power value of the whole point at 11 hours of a day.The results show that the GWPA-BP neural network photovoltaic power forecasting model has faster convergence speed and higher prediction accuracy,it can be very good to the short-term photovoltaic power forecasting and widely applied in the power system of high accuracy requirements.