Study on Dynamical SVM Models for Extended Short-Term Load Forecasting in Electricity Market

Daily load forecasting model does not update its input variables with newly obtained load and climatic information to dynamically trace the latest variation of load.The extended short-term load forecasting can improve the accuracy by appending the latest information to the models and forecasting the following hourly loads.Support Vector Machines (SVM) is applied to build a series of dynamical forecasting models with newly obtained load information as part of its input variables.Then these models are used for rolling forecasting the rest load of the day.Experiment results show that the forecasting errors with the dynamical models are at least 1/3 lower than the compared methods.