Fisher Information Based Climate Variables Modeling Method For Short-Term Load Forecasting

In the short-term load forecasting,the correct processing of meteorological factors directly affects the accuracy of forecasting;this paper analyzes the periodical law of the load of the power system and its relationship with the smart grid environment that can provide a large amount of real-time load data and meteorological data.Based on the relationship between the real-time meteorological factors and the daily characteristic meteorological factors,a Fisher information modeling method for meteorological factors in short-term load forecasting was proposed,and numerical experiments were performed in combination with specific examples.Excavating suitable weather factor processing methods to further improve prediction accuracy has always been a key and difficult point in short-term load forecasting.Fisher information theory provides a new way for us to solve such problems.This method is used to solve the modeling problem of real-time meteorological factors in short-term load forecasting under big data environment.For one or more meteorological variables,this paper first solves the computational problem of one-dimensional or multidimensional Fisher information.Based on this,the modeling method and forecasting model of meteorological factors based on Fisher information are given.In recent years,artificial intelligence models have become increasingly popular in short-term load forecasting.Support vector machines(SVMs)have been most widely used in short-term load forecasting due to their better generalization performance and higher prediction accuracy.Combinin g the characteristics of load forecasting based on meteorological factors,this paper replaces the slack variable with the square of training error of support vector machine(SVM),introduces the least square support vector machine(LSSVM),and solves the problems of nonlinearity and local minimum of SVM.Then the LSSVM was used to verify the new model constructed using the Fisher information-based meteorological factor modeling method.In order to verify the general applicability of the Fisher information modeling method for meteorological factors in short-term load forecasting,the BP neural network was used to verify the new model.The actual test results show that whether it is a least squares support vector machine or a BP neural network,the prediction model established by using the meteorological factor modeling method based on Fisher information can obtain more accurate prediction results and solve the problem of long-term short-term load forecasting.The subjective arbitrariness of the meteorological factors has led to a unified treatment of the problem.