Research On Forecast Of Photovoltaic Output Based On Extreme Learning Machine

Today,as the country’s environmental pollution continues to increase,the energy structure is facing tremendous changes,and renewable energy has been vigorously developed.Photovoltaic power generation is favored by countries all over the world due to its safety,cleanness,high efficiency,pollution-free,and noise-free advantages.The installed capacity has steadily increased.However,due to the intermittent and random nature of the active output of photovoltaic power sources,large-scale grid connection of photovoltaic power stations will inevitably pose a threat to the stability of the power system.In order to ensure the safety and stability of its grid connection process and make full use of the advantages of photovoltaic power generation,this article has conducted in-depth research around photovoltaic power generation power prediction technology.First of all,this article summarizes the domestic and foreign experience and methods of photovoltaic power generation power prediction technology,explains the composition of photovoltaic power generation systems and their main mechanism of action,the classification of photovoltaic power generation systems,and analyzes the impact of photovoltaic power generation power prediction technology based on field data.Factors,analyze and judge the impact of different factors on the forecast of photovoltaic power generation.Then,several basic theories of photovoltaic power prediction are introduced,which lays a theoretical foundation for subsequent power prediction,and the evaluation criteria of photovoltaic power prediction errors are analyzed.In order to verify that the combined prediction method is more effective than a single algorithm,and is more excellent in accuracy,timeliness,etc.,a targeted photovoltaic output prediction model based on the gray extreme learning machine combined model is proposed,which effectively integrates gray prediction and extreme learning machine prediction The characteristics of the model,give play to the advantages of both,to solve the shortcomings and problems of traditional single model power generation.In order to effectively control the error problem of the hidden layer bias and input weights of the ELM network under the randomly given action,based on the analysis of the advantages and disadvantages of the existing models,the chaotic particle swarm optimization and the genetic particle swarm optimization extreme learning are proposed.The photovoltaic power generation model of the machine,at the same time,by improving the connection weights and thresholds,it evaluates the scientific and reasonable output weight matrix,which effectively reduces the error under the influence of random parameters.The experimental comparison and analysis of the prediction accuracy and error of the algorithm before and after the optimization show that the photovoltaic power generation power prediction method established by the improved particle swarm optimization optimization extreme learning machine has higher accuracy and effectively predicts the operating status of the power station.