A new method for short-term electricity load forecasting

Accurate forecasting of short-term electricity load is an important issue in the electricity industry. This paper proposes a new forecasting model by integrating the support vector machines (SVMs) forecasting technique and rough sets (RSs) with reduced attributes using evolutionary algorithms (EAs). Simulation results show that this new model can improve the prediction accuracy, speed the convergence and require less computational effort in comparison with another two methods, namely the traditional SVM model and a model combining the SVMs and simulated annealing algorithms (SVMSA). This improvement is related to fact that the RS techniques can reduce the SVM input variables and improve the convergence.