Research On Power Load Forecasting Based On Online Adaptive Least Squares Support Vector Machine

In recent years,a large number of new energy power has been connected to the grid,which has increased the randomness and intermittentness of power supply on the power supply side.Coupled with the development of intelligent power grids,various control applications urgently need power load forecasting data as basic research data,so power load forecasting is put forward higher requirements.Load forecasting affects many decisions of the power system,such as economic dispatch,automatic power generation control,safety assessment,maintenance dispatch and energy commercialization.Accurate power load forecasting can economically and rationally arrange power generation plans and power dispatch,and plays an important role in effectively reducing power generation costs.Electric power load forecasting requires deep mining of load and load-related characteristic variable data based on machine learning methods,and predicting load characteristics by establishing a data-driven model.According to the data-driven principle of machine learning,a power load forecasting method based on partial least squares(PLS)and online adaptive least squares support vector machine(OALSSVM)is proposed and compared with other power load forecasting methods.Validated by actual power load data in a certain area,the variable characteristic calculation method based on PLS can effectively affect the load and reduce the complexity of the model.The OALSSVM load model can accurately predict the load along with the shift of working conditions,and has a high generalization ability and engineering application value.This article first analyzes the current research status of power load forecasting at home and abroad,and points out the existing problems in current power load forecasting research.On this basis,the main content and methods of this subject research and the significance of the research are put forward.Then the basic principles and basic ideas of power load forecasting and variable feature extraction are described,and the commonly used power load forecasting methods,the algorithm principles of PLS and OALSSVM are introduced.Then,the contribution value of the independent variable to the dependent variable is calculated based on the PLS method.The characteristic variable selection of the electric load is realized by comparing the magnitude of the contribution value.Finally,a power load forecasting method based on PLS-OALSSVM is proposed,and the actual power load data of a certain area is used to verify the method proposed in the article.On this basis,a power load forecasting system was built.