Power Load Forecasting Based On Intelligent Optimization Model

With the rapid development of smart grid technology,the requirements for the speed and accuracy of power load forecasting are increasing day by day.Power load forecasting is a method used by power generation companies to estimate the total power required to provide customers.Load forecasting is closely related to saving electricity,maintenance and repair of generator units,power grid dispatching and ensuring the stability of production and domestic electricity.It is the key task of various relevant units such as power system planning and operation.The core task of load forecasting is to infer the future development trend according to the change law of previous load data.This paper optimizes and improves the traditional power load forecasting method,and proves the superiority of the optimized method through the comparison of evaluation indexes.The main research contents are as follows:This dissertation focused on optimizing the parameters of support vector machine model and improved the performance of support vector machine model.Firstly,the principle of support vector machine was introduced,then the optimal parameter combination of support vector machine prediction model was found through the traditional grid search method,and the prediction model was constructed with it,then trained,and finally predicted.Then,particle swarm optimization,an intelligent algorithm,was proposed to optimize the support vector machine model.Through comparative experimental analysis,it was concluded that better prediction effect was achieved.Then,in view of the decline of search ability in the later iteration of particle swarm optimization algorithm,simulated annealing algorithm with strong local search ability was introduced to improve it to form complementary advantages.Finally,comparative experiments showed that the optimization effect of the fused algorithm was better,and the convergence speed was improved.Finally,aimed at the disadvantage of premature convergence of Gray Wolf algorithm in finding the optimal parameter combination of support vector machine load forecasting model,differential evolution algorithm was used to improve it.The improved gray wolf algorithm expanded the search range,and then used it to find the optimal parameter combination of support vector machine load forecasting model,used it to build the forecasting model,then trained,and finally forecasted.The simulation results showed that the load forecasting model based on Improved Grey Wolf algorithm and support vector machine had better accuracy and rapidity.In this dissertation,date,temperature,month,holiday and working day were used as the input and load as the output of the forecasting model.The data of one year in a certain place was selected to form the training set,and the data of January in the second year was used as the test set The comparative experiments showed that the average relative error of the load forecasting model based on PSO-SVM was 2.343%,which was further reduced compared with 2.779%of the load forecasting model based on SVM.The average relative error of the load forecasting model based on SAPSOSVM was 1.418%,which was lower than 2.343%of the load forecasting model based on PSO-SVM.The average relative error of the load forecasting model based on DE-GWO-SVM was 2.086%,which was further reduced compared with 2.569%of GWO-SVM model and 2.779% of SVM model.