Improved Markov Residual Error to Long-Medium Power Load Forecast Based on SVM Method

The characteristics of small sample, stochastic growth and nonlinear wave are often combined with long-medium power load forecast series; SVM model could reflect the relationship between growing characteristics and nonlinear characteristics to the series effectively and make fitting calculation, on the other hand, Markov could well reflect randomness that produced by the system involve with many complex factors. Through establishment of a forecast model based on SVM algorithm, the series of historical load variables is rolling forecasted; an improved Markov error correction algorithm is introduced to modify the values forecasted by SVM, in order to make the increase of total forecasting precision to a maximum extent, a transfer matrix that make the forecast values to high stability and high accuracy is obtained. It is proved that the presented forecast method is superior obviously to traditional methods through empirical study, and it can be used generally.