温斯特线性与季节性指数平滑法在电力负荷预测中的应用及改进

由于电力负荷以年为单位周期性波动,且呈现出逐年上升的趋势,因此,适合于用温斯特线性与季节性指数平滑法进行预测. 但由于此方法对于负荷数据中的异常值极其敏感,很有可能导致预测结果与实际电力负荷趋势相反. 针对这一问题,对原模型作了相应的改进,使其预测结果更加符合电力负荷的总体趋势,不致于因个别异常值的存在而得出与事实相反的结论,并能进一步减少模型误差.

Application and Improvement of Method of Wenshite's Linear and Seasonal Exponential Smoothing to Electric Load Forecasting

In terms of its' characteristics of seasonal vibration and increasing tendency year by year, electric load is suitable to be forecast by the method of Wenshite's linear and seasonal exponential smoothing. In applications, it is also possible for the forecasting method to draw contrary conclusions to the real tendency caused by its' high sensitivity to abnormal data. A new measure is put forward in this paper to improve the method for avoiding the contrary conclusions. It shows that the impact from abnormal data widely declines and the forecasting result more depends on the general tendency of electric load.