Research And Application Of Wind Power Forecasting Technology Based On Deep Learning

As people’s demand for energy continues to increase,fossil resources such as coal,natural gas,and oil are becoming less and less stored.The combustion of these energy sources will produce a large amount of harmful substances,causing environmental pollution.As one of the renewable energy sources,wind energy is widely used all over the world due to its wide distribution,large reserves,green and clean,and it is an important part of large-scale grid integration and green.In response to the 3060 target of carbon peak and carbon neutrality,it is important to reduce carbon emissions through vigorously developing new energy sources.But,wind power is random,fluctuating,anti-peak,and wind power forecasting has low accuracy problems,which is harmful to the power system.So,wind power is paid attention to forecasting and improving the accuracy of wind power forecasting.Due to the large difference in wind power series,direct use in data training may cause excessive errors.Therefore,this article first uses clustering method to cluster wind power series.In order to keep the power grid system operated safely and stably,dividing typical wind power output scenarios,and divides wind power output Scenario-based,accurate seasonal prediction of typical scenarios of wind power output is necessary.First,perform data preprocessing on wind power data to correct abnormal or missing data.Then,divide the typical scenarios of wind power output.GMM clustering and k-means clustering are used to predict typical wind power output scenarios in spring to verify the effectiveness of GMM clustering.Cluster analysis of typical scenarios of seasonal wind power output is done through the GMM clustering model to generate cluster center curves for the four seasons of spring,summer,autumn,and winter.Lay the foundation for the prediction of typical scenarios of wind power output.Finally,predict the typical clustering scenes of spring,summer,autumn,and winter seasons.When using deep learning network to predict,there will be a problem of determining hyper-parameters.This article uses an improved particle swarm optimization algorithm(IPSO)to perform hyper-parameters.Then use the optimized deep learning neural network to predict the wind power.The IPSO-LSTM model is compared with the LSTM model,the PSO-LSTM model,and the GA-LSTM model.The scene chooses four seasons of clustering center curves.Aiming at the proposed IPSO-GRU model,compare it with the GRU model and PSO-GRU model to verify the accuracy of the model.The experimental results show that both the IPSO-LSTM model and the IPSO-GRU model have high prediction accuracy,and the IPSO-GRU model has the highest prediction accuracy.