"Mechanism Model+Identification Model" Strategy And Its Application To Electric Short-term Load Forecasting

Short-term load forecasting of electric power systems is a conventional important problem for gaining their benefits in economy and stability. At present, the main studies focus on employing the advanced intelligent forecasting techniques and synthesizing optimally the multiple models'loads forecasted. However, the chief way to improve the accuracy of forecasting must be the study on strategies of forecasting.After cogitating to combination forecasting strategy, we proposed the"mechanism model + identification model"forecasting strategy which can comprehensively consider the structure information and historical behavior of complex systems. Firstly, the regressed loads were gotten by multiple linear regression from the daily five weather factors of the latest continuous three days; Secondly, the remainders of subtracting regressed loads from record loads were forecasted by identification models; Finally, the strategy of behavior's forecasting after multi-models estimations advocated by DDDAS was used to synthesize the multi-loads of forecasting. Simultaneously, the ensemble technology of numerical weather forecasting was transplanted into short-term load forecasting in order to solve the initial values'sensitivity caused by matrix singularity in numerical calculation.The main achievements are: (1)"Mechanism model + identification model"strategy enhances the stability of the forecasting results, as the remainders are less than five percent of record loads, and the stationarity of the remainders series is improved observably. (2) Ensemble prediction can not only enhance the accuracy of the loads forecasted, but also realize their probabilistic forecasting. (3) the loads forecasted by the strategy of behavior's forecasting after multi-models estimations are primarily better than those by the practical method of the combination forecasting.The research suggests that using"mechanism model + identification model"strategy and the strategy of behavior's forecasting after multi-models estimations advocated by DDDAS to forecast short-term load is a beneficial new forecasting strategy. These methods can obviously improve the accuracy of the forecasting loads and realize their probabilistic forecasting.