Short-term Load Forecasting Approach with SVM and Similar Days Based on United Data Mining Technology

In view of the defect of support vector machine (SVM) when used for short-term load forecast such as large data amount, long training time and too much redundant information, the paper proposes a SVM forecasting model with similar days based on united data mining technology. Taking advantage of the data mining technology in processing large data and eliminating redundant information, uses a combined approach of several mehods and chooses some historical days whose characteristics is highly similar to forecasted day with the consideration of various factors such as date characteristics, weather conditions and range of average load, and then make the new data sequence as the training sample of SVM. With the comprehensively consideration of various factors, it not only simplifies the input dimension and calculation of the model, but also intensifies the regularity of historical data, thus is more convenient for application. The numerical example shows the effective and available of the proposed method.