Complex seasonal circular block bootstrap for electricity load forecasting

Abstract

We propose the Complex Seasonal Circular Block Bootstrap (XSCBB), a variation of seasonal (circular) block bootstrap that caters for multiple seasonality components in a time series. Electricity consumption (load) prediction is important to balance the supply and load demand, to plan facilities construction and maintenance, to plan distribution, and avoid outages or excess loss. We apply the XSCBB method parametrically to calculate the prediction interval of future electricity consumption given a relatively small amount of historical sample points using the composite ARMA$\left(p,q\right)$– GARCH$\left(r,s\right)$ model.