1) Short-term load forecasting of power system is a
challenging research field. An effective way to improve the
prediction accuracy is to combine actual situation, analyze the
load characteristics in detail, to select the feature vector
accurately and a method with good performance.
2) The problems of small sample, nonlinear and local
minima can be well resolved by support vector machine that
has good generalization ability, and the LIBSVM algorithm
has fast computing speed and high forecast accuracy.