The Application Of Combinational Forecasting Based On Support Vector Machine In Mid-long Term Load Forecasting

The forecasting to mid-long term load can provide important evidence to the power planning. The accurate load forecasting can improve the economics and reliability of power system operation. The load is related with a variety of factors, there is a complex nonlinear relationship between the load and the factors. The forecasting to mid-long term load is important because it can provide important evidence to the power planning.As a new machine learning algorithm, support vector machine (SVM) can solve some practical issues, such as small sample, nonlinear, high dimension and local minimum points, etc. Compared with the short-term electric load data, the long-term data has the characteristics of small samples. SVM has the advantages to the small samples that other models can not be compared with, and the SVM regression method has good capability of fitting and extrapolation.Traditional forecast techniques apply a single forecaster to carry out the task. However, this forecaster might not be the best for all situations or databases. A combinational model on the basis of Support Vector Machine (SVM) theory is proposed in this paper. During the process of the forecast, several single forecasting methods such as trend prediction model, exponent model, non-linear regression model, improved grey predictive model and improved grey verhulst predictive model, are used to form a model group, and then the fitted results by different traditional predictive models in time sequence act as the input of the support vector machine regression (SVMR) model, then by relative SVMR approach based on known input and output samples, we can obtain the test model. In the paper, the procedure of the combinational prediction on transformer faults based on SVMR is discussed in details. The example on load data has proven that the proposed model can give good results on both the fitting to the known data in time sequence and the extrapolation to the data to be predicted. Moreover, compared with other predictive approaches, both single model and other combinational model, the proposed combinational forecasting model has higher prediction accuracy.