Research On The Short-term Load Forecasting Based On The Improved Support Vector Machine

Load forecasting is one of the most important part of the stable operation of the electric power system, and it is attached great importance by researchers and enterprises. Short-term load forecasting means a lot to the safe, stable and economic operation of the power grid. Firstly, the precise short-term load forecasting helps power enterprises to reasonably arrange the start-stop of generating sets, which ensures the stability of the grid operation. Secondly, accurate forecasting results help to reduce redundant back-up capacity and provide basis for the maintenance planning of generating sets. Furthermore, short-term load forecasting is both a crucial basis for the energy-saving dispatching of the power grid and a foundation for the safe and reliable supply of the electric power. Finally, the management of short-term load forecasting involves many sections of power grid enterprises, such as planning, dispatching and promotion departments, which has a tight connection to the negative factors of operation accidents and property losses. Therefore, it is of great significance to further study short-term forecasting methods and models.This paper puts forward a review on the development trend of domestic and international study on load forecasting methods, elaborates on the basic principles such as short-term load forecasting, data mining and Support Vector Machine(SVM), which establishes a foundation for the study on the improved short-term load forecasting algorithm. Combine with the SVM model, the study verifies the applicability and superiority of SVM theory on the short-term forecasting area through empirical studies. Furthermore, the study improves the SVM model through Simulated Annealing algorithm, which makes the SVM model possess "memory" ability when selecting parameters, and the whole parameters enhance the ability of adjusting optimization directions. Through empirical and comparative analysis, it is proved that the improved SVM model has better forecasting accuracy and effect, which provides substantial theory and practical basis for further improving the load forecasting standard.