Forecasting Daily Peak Demand Using GAMs

In Chap. 2 we introduced GAM models, while in Chap. 3 we extended the GAM modelling framework to cover probabilistic demand forecasting using GAMLSS and QGAM models. In the applications presented in those chapters we were forecasting half-hourly demand using covariates that were defined either at the same time resolution as the demand (e.g., half-hourly temperatures) or less frequently (e.g., the day of the week). Here we consider the problem of forecasting the size and time of the daily peak demand which are, of course, observed on a daily basis, using covariates that are observed more frequently.