Point and interval forecasts of electricity demand with Reg-SARMA models

Abstract This paper deals especially with a two-stage approach to forecasting hourly electricity demand by using linear regression model serially correlated residuals. Firstly, ordinary least squares are applied estimate based on purely deterministic predictors (essentially, polynomials in time and calendar dummy variables). In the case wherein residuals not white noise series, SARMA (seasonal autoregressive moving average) process is estimated After examining vast set of potential representations, stationary invertible associated smaller Akaike information criterion Ljung–Box statistic selected. Secondly, two sets instrumental added current model: first plus errors chosen process. The new again squares, but taking advantage fact that regressors eliminate serial correlation. Practical issues points interval illustrated reference nine-day ahead prediction performance for short-term electric loads Italy.