Short-term Power Load Forecasting Based On Information Granulation And Weighted SVM

Power system short-term load forecasting as an important means to ensure the safety and economic operation of the system,accurate load forecasting can effectively improve the utilization rate of electricity and reduce unnecessary losses.In the short-term load forecasting,most of the research is to predict the specific load value of the next moment,but the load change trend and the range of forecasting less research,so this paper on the basis of the traditional support vector machine,Short-Term Load Forecasting Algorithm Based on Support Vector Machine with Information Granulation and Similarity Weights.To this end,this article did a bit of work:(1)using weighting SVM based on similarity weight.Support vector machine(SVM)is one of the powerful tools in machine learning field,and it has excellent nonlinear learning and prediction ability for dealing with regression prediction problem.However,the impact of the standard support vector machine on the difference of the sample can not be eliminated.In this paper,the similarity calculation is introduced,and the main factors of short-term load forecasting are quantified to form the daily eigenvector.The similarity coefficient method is used to calculate the historical day The similarity of the forecast day and the similarity as the weight of the importance of each sample data,so as not only consider the factors that affect the load forecast,but also improve the accuracy of the algorithm.(2)using SVM based on fuzzy information granulation.Since most of the load forecasting methods study only a single real-time forecast,there are few studies on the trend of load change and fluctuation space.In this paper,we use a reliable range solution instead of a single value solution from a more practical point of view,combining the information granulation theory with SVM,selecting triangular fuzzy particles and the size of each window as a window,and comparing the sample data And then use the SVM regression model to train and predict the data after granulation to obtain the range of load changes in the next window period.Finally,through the simulation experiment,the load data of regional power grid is used to test the short-term load forecasting model.The results show that the short-term load forecasting model based on information pelletizing SVM with small cost and high efficiency can effectively predict short-Load change trend and change space.