Electrical load forecasting using time series model

Abstract

Many forecast models such as regression, exponential smoothing method, fuzzy regression, multilayer perception and extreme learning machine have been proposed to forecast daily electrical load. But some of the models do not incorporate the autocorrelation structure and they are not easy to interpret the forecast results. In this paper, we introduced transfer function and intervention model using discomfort index, sensory temperature index as input time series and seasonal eect, sandwich day(the day is between two holidays) eect as intervention. This model allows us to interpret predictive value and to forecast more accurately. This model might be quite useful to save power cost and to supply electricity smoothly