Managers and researchers have put more emphasis on electrical power system forecasting to obtain effective management in electrical power system. However, enhancing prediction accuracy is not only a highly challenging task, but also a concerned problem in electrical power system. Traditional single algorithms usually ignore the significance of parameter optimization and data preprocessing, which always leads to poor results. Thus, in this paper a novel hybrid forecasting system was successfully developed, including four modules: data preprocessing module, optimization module, forecasting module and evaluation module. In this system, a signal processing approach is employed to decompose, reconstruct, identify and mine the primary characteristics of electrical power system time series in data preprocessing module. Moreover, to achieve high accuracy and overcome the drawbacks of single models, optimization algorithms are also employed to optimize the parameters of these individual models in the optimization and forecasting modules. Finally, evaluation module including hypothesis testing, evaluation criteria and case studies is introduced to make a comprehensive evaluation for this system. Experimental results showed that the hybrid system not only can be able to satisfactorily approximate the actual value, but also be regarded as an effective and simple tool adopted in smart grids. (C) 2018 Elsevier Ltd. All rights reserved.