Online daily load forecasting based on support vector machines

In this paper a new online daily load forecasting (ODLF) approach is proposed for short term electrical load forecasting in practice. It incorporates the load forecasting algorithm based on the support vector machines (SVMs) with OPC specifications and becomes a standard online load forecasting component. By carefully selecting the load features in the electric power system and using the improved sequential minimal optimization algorithm, this algorithm can accurately predict the next day´s load trend online. Especially the SVMs-based ODLF approach showed satisfied performance, such as powerful regression ability, acceptable predict accurateness and perfect foundation in theory. As an independent functional modular, the designed ODLF component can be used in any OPC-compatible environment for real-time load forecasting in distributed electrical system.