Prediction Of Photovoltaic Power Generation Based On PSO-BP Neural Network

With the increasing proportion of photovoltaic power generation globally,the prediction of accuracy of photovoltaic power generation will directly affect the strategy and economy of photovoltaic power grid connection.However,due to the fluctuation and intermittency of photovoltaic power generation,the conventional prediction methods often differ greatly from the actual results,and photovoltaic cells have shortcomings such as wide footprint and low conversion efficiency.In addition,if the photovoltaic power generation is Be applied to the main power grid without treatment,it is easy to cause a strong impact on the main power grid,which is extremely detrimental to the stable operation of the power grid.Therefore,the accurate prediction of photovoltaic power generation is an important prerequisite to reduce the adverse impact on the power grid.It will not only cause light abandonment and reduce the generation income,but also bring problems to the grid connection and affect the overall stability of the power grid if he generation power cannot be predicted more accurately.Based on the comparison between the new energy and the traditional energy,this document analyzes the principle of photovoltaic power generation and the current commonly by using power prediction methods.Considering the good effect of artificial intelligence algorithm on nonlinear mapping fitting,the particle swarm optimization BP neural network(PSO-BP algorithm)is adopted as the prediction algorithm.In view of the volatility and randomness of photovoltaic power generation,the traditional PSO-BP algorithm is improved,and the iterative formula of particle swarm optimization algorithm is adjusted and the parameters are adjusted according to the specific example.In terms of the original data,the data of the self-built photovoltaic power station of the University of Oregon in the United States were used as the original sample data,and the data were selected and normalized to reduce the errors caused by the defects of the original data.In terms of the algorithm network construction,the method of using similar days as the forecast benchmark is adopted,and similar days are found by calculating the Euclidean distance between the measured day and the weather data of the sample database.Kolmogorov theorem and Try-error-try method are used to determine the number of nodes and layers in hidden layer of network.In order to ensure the reliability of the network model,the method of random cross grouping has been used in this document to carry out grouping modeling,and select the group with the smallest error as the final prediction model.In the document,an algorithm model for predicting photovoltaic power generation by using PSO-BP algorithm is established,and the simulation on the same day is verified by using traditional BP neural network and PSO-BP algorithm.The simulation results show that the overall fitting relative error of PSO-BP is 5%,which is better than the 9% of traditional BP neural network.Meanwhile,PSO-BP algorithm can also quickly obtain the fitting interval to achieve fast iterative solution.The simulation results prove the feasibility of PSO-BP algorithm applied to the power prediction of photovoltaic generation.After the simulation has been completed,the error analysis of the simulation prediction results is carried out.In view of the large error of the time node before sunset,the main reason of the error is discussed,and the research direction of improving two algorithm models is proposed according to the analysis results.