Short-term Photovoltaic Power Prediction Technology Based On Artificial Intelligence

With the aggravation of environmental pollution and the adjustment of energy structure in China,renewable energy has been vigorously developed.Photovoltaic power generation is favored by countries all over the world because of its advantages of safety,cleanness,high efficiency,pollution-free and noise-free.Among them,the installed capacity of photovoltaic power generation is steadily increasing.However,the intermittent and random nature of the photovoltaic power output makes the large-scale grid connection threaten the stability of the power system.In order to ensure the safety and stability of the grid connection,and make full use of the advantages of photovoltaic power generation,this thesis will conduct an in-depth study around the photovoltaic power prediction technology.Firstly,this thesis summarizes the previous experience of photovoltaic power generation technology,expounds the basic principle of photovoltaic power generation,and analyzes the influencing factors of photovoltaic power generation based on field data,compares and analyzes the influence degree of each factor on the photovoltaic power generation through the similarity coefficient.Secondly,in order to improve the data quality of photovoltaic power generation prediction and eliminate the effects of weather types,the thesis uses the improved gray correlation theory to select similar correlation data from the original data set,and normalize the data to eliminate the impact of data size due to different data dimensions.On this basis,the thesis analyzes the mathematical principles of back-propagation(BP)neural network,decision regression tree,support vector regression algorithm and gradient lifting tree.The thesis compares and analyzes the accuracy and error of the four algorithms through experiments.The experimental results show that the BP neural network algorithm has higher accuracy in the photovoltaic power generation prediction models based on artificial intelligence algorithm,but it is easy to over-fitting in the process of training the model,which affects the final prediction effect.To this end,this thesis uses Bayesian neural network to solve the limitations of the above artificial intelligence algorithm,and explains the working principle of Bayesian neural network through mathematical theory.Bayesian neural networks have stronger generalization ability and stability,and can effectively prevent over-fitting.The thesis compares and analyzes the prediction accuracy and error of the algorithm before and after optimization,and it is shown that the accuracy of predicting photovoltaic power based on Bayesian neural network is higher.