Power System Short-term Load Forecasting Based On Support Vector Machine (svm)

Short-term load forecasting is an important day-to-day work of power system scheduling department. It’s the main basis for making the power purchase plan and operation mode. Its prediction accuracy will directly affect the security and economy of power system. With the depth of the power system market-oriented, short-term load forecasting accuracy and speed have become increasingly demanding.In this paper, an area of historical load and weather data analysis, to explore a load variation of the main factors that affect the accuracy of load forecasting.The support vector machine has excellent non-linear learning and forecasting features, this article will be used for short-term load forecasting.The core of the support vector machine is a kernel function, kernel function selected by the model, nonlinear changes in the data space and the high-dimensional space is different, making the prediction results and the accuracy with certain differences. This paper constructs a short-term load forecasting model based on a hybrid kernel of support vector machine, the model will be different kernel function weighted, making it both a combination of kernel functions of the global nuclear and partial nuclear superiority. The actual power load forecast simulation and testing proved that the prediction accuracy is better than the support vector machine based on a single kernel function.For a single prediction model is not fully reflect the variation of power load and information, this paper construct a combination prediction model with the SVM (Support Vector Machine) and AM-NN. The two sub-models combines together by the optimal weighting coefficients. The actual power load forecasting simulation and testing confirmed that the proposed combination forecasting model can effectively improve the prediction accuracy.