Short Term Power Laod Forecasting Based On Least Squares Support Vector Machines

The continuous improvement of economy and technology make people's requirement for electricity consumption become higher and higher.And the power system short-term load forecasting technology is of great significance to ensure the safety,stability and economic operation of the power system.No matter in the traditional power grid environment or in the new environment of smart grid,high quality short-term load forecasting is becoming more and more urgent.In this paper,a method of power system short-term load forecasting based on least squares support vector machine is proposed.In short-term load forecasting,the characteristic of power load is the basis for constructing load-forecasting model.In this paper,we first fully understand the characteristics of power load by plotting a region's power load distribution of daily maximum load,daily minimum load,daily peak-valley difference and daily load rate in 2016.Based on this,we establish a simple regression model.The influence of the meteorological factors such as the highest temperature,the lowest temperature,the average temperature,the relative humidity and the rainfall on the electric load is obtained.Then,the abnormal data in historical load data is revised and the influencing factors of power load forecasting is normalized.Secondly,based on load characteristics analysis,input sample feature selection and historical data preprocessing,a least squares support vector machine prediction model is established.Least squares support vector machine is a new machine learning method based on statistical learning theory.it has the outstanding advantages of strong generalization ability,global optimum and fast computation speed,and has been successfully applied to short-term load forecasting.Finally,in order to further improve the prediction accuracy,aiming at the shortcomings of the current least square support vector machine parameter selection method,the improved particle swarm optimization algorithm is used to select the least square support vector machine.The improved particle swarm optimization algorithm can guide the selection of initial population according to the information of population diversity,avoiding the premature convergence problem.The prediction results are evaluated by the average relative error of the prediction results and compared with the standard least square support vector machine prediction model.Through the example analysis,it is verified that the improved particle swarm least squares support vector machine model has excellent generalization and global optimization ability.It has high accuracy,good convergence and faster training speed.