Research On Power Load Forecasting Based On Artificial Intelligence

Energy is the foundation dynamic for the progress of mankind,and energy development is a hot issue of global concern.To promote efficient,clean,and green energy development,foster the vision of a community with a shared future for mankind,and promote all-around green economic and social transformation.This requires us to respond aggressively to a new national energy security strategy.Electric power energy is one of the significant energy sources for national development,which not only relates to all aspects of our daily life but also directly affects the economic development of the country.At present,during a fast year of economic development in our country,good planning of the electric power system can guarantee the stable operation of the power grid and provide a high-quality power energy supply.The deepening of power forecasting research and the development of smart meters,smart sensors,and other equipment has laid a solid foundation for power system forecasting.In the actual research work,the power load has the characteristics of time sequence,periodicity,random fluctuation,continuity,and so on.Accurate power system prediction is an important direction of load research.According to the relevant requirements of the power system,this thesis has summarized the shortcomings of the traditional power load forecasting methods,and excavated the advantages of the traditional algorithms.From the perspective of algorithm complementarity,the prediction model has been combined and optimized,and a prediction model based on fuzzy C-means clustering and improved local weighted linear regression is built.The load data with the same electricity consumption behavior have been clustered to make the load data more targeted.Introducing K nearest neighbor algorithm into the local weighted linear regression prediction model can save computation and improve prediction accuracy.The comparison experiment of the combined model has been designed.The experimental results show that the proposed method has better generalization ability,and the prediction accuracy is greatly improved compared with the model before the combined optimization.To further improve the prediction accuracy,the clustering algorithm has been improved,and the kernel fuzzy C-means clustering method based on particle swarm optimization has been proposed to cluster the data sets,which made up the problems of poor clustering quality and sensitive clustering center of the original algorithm.The change in external influencing factors will also affect the prediction accuracy to some extent.Considering the influence of time and weather factors,the time attention mechanism and channel attention mechanism have been introduced into the time domain convolutional network model.Besides,feature selection has been carried out by the Pearson correlation coefficient,and high-quality influence factors have been taken as inputs.The experimental results show that the proposed method has good fitting ability and stability.