Short Term Load Interval Prediction Of Power System Based On Gauss Kernel Density Estimation

The short-term load forecasting of power systems is based on the past and present load values and related influencing factors to predict future values.It plays an important role in the economic stability of the power system.With the rapid development of modern artificial intelligence methods and applied to power load forecasting,better deterministic load forecasting results can be obtained.Due to the rapid development of intelligent electric grid,factors that affect the accuracy of power load forecasts include,in additio n to historical load and meteorological information,various uncertainties in the power system.However,the results of the deterministic forecasting method do not reflect the uncertainty of the demand.Interval forecasting can meet this objective requirement.Power system decision-makers can make better decisions to better understand the possible uncertainties of future loads based on the results of interval forecasting when working on production planning,system safety analysis,etc.This paper analyzes the load characteristics to identify the influence of the load itself and the related factors.Least squares support vector machine model is used for prediction.However,the two parameters in the model have a great influence on the model.In this paper,particle swarm optimization algorithm is used to optimize the model parameters.A Least Square Support Vector Machine(SVM)model based on particle swarm optimization(PSO)was established to optimize the parameters of the model.The deterministic prediction of the future short-term load was performed.The average relative error was used as the evaluation criterion to validate the accuracy of the algorithm.For the uncertainty of power load changes,based on short-term load deterministic prediction,this paper proposes a short-term load interval prediction method based on Gaussian kernel density estimation with optimal window width.By statistically analyzing the relative error characteristics of historical load forecasting,this method first uses the non-parametric kernel density estimation method to select the Gaussian kernel function and the optimal window width,and models the relative probability error probability density function in each load partition.Then combine the deterministic prediction results into the probability density function of the future load values.Finally,use the cubic spline interpolation method to calculate the load forecasting interval under different confidence levels of the deterministic predictive value of the load.In this paper,the interval coverage rate and the average width of forecast interval are taken as the evaluation index of the output interval.The example proves that the proposed short-term load interval prediction method is feasible.