Research On A Regional Power Grid Forecasting Based On Support Vector Machine

ABSTRACT:Electricity demand forecasting is an important part of power system planning. This paper mainly for long-term maximum power load forecasting. Its characteristics are less historical data, the load being affected by the economic and social impact of large uncertainties. Accurate load forecasting is conducive to improve the security and stability of the power system operation, effectively reduces the cost of power generation, ensures the electricity demand, and enhances supply reliability, thereby enhances the economic and social benefits for power system.This article describes the purpose and significance of the power system load forecasting, load forecast for the current situation at home and abroad were reviewed Introduces the basic principles of load forecasting, analysis of the advantages and disadvantages of each method. Describes the classification and characteristics of the power load, given the requirements of the power load forecasting model and error indicators. Meanwhile SVM literature review, pointed out the advantages of SVM and problems. And the factors affecting the demand for electricity in Shanxi Province were analyzed.This paper analyzes the characteristics of the power load forecasting based on support vector regression algorithm Shanxi Power Grid monthly maximum power load forecasting. The basic principles of support vector machine is introduced. The model and algorithm flow chart based on this method are presented. The realization of this algorithm is based on MATLAB program design. Comparing with other methods, the actual example shows that the forecasting model is in line with the characteristics of medium and long-term load forecasting, it is feasible.Then the analysis of various parameters of support vector regression machine has a great influence on the basis of its performance, combined with the monthly maximum power load characteristics, proposed support vector load forecasting model using particle swarm optimization, and organize monthly load data by weigh recent law.The optimization principles and flow chart of the modified model are presented. Through analysis of practical examples, and standard support vector regression method to predict the results were compared, proved PSO-SVR forecasting model has predicted a high precision, low computational advantages.Meanwhile, the proposed PSO-SVR improvement measures to increase the inertia weight factor, and made a three-point smoothing method to optimize data to make the model more perfect. Given the improved algorithm flowchart, through monthly load forecasting Shanxi proved the PSO-SVR improved forecasts more accurate. On the basis of the factors affecting demand for electricity on the previous analysis. For the annual load of features, sorting the input data, the use of well-established model to predict the maximum power load for future years.Finally, commonly used at home and abroad based on support vector machine algorithm. Analysis of the grid search method to optimize the characteristics of support vector regression, optimizing the use of support vector machine based grid search method CMSVM software Shanxi power load forecasting research. Summarize the key issues and make load forecasting based on support vector machines that need attention, and make recommendations.