|
Abstract:
|
With the advent of wholesale electricity markets there has been renewed focus on intraday forecasting of both electricity demand (or load) and spot prices. This presentation discusses a variety of approaches for modeling both electricity load and spot prices. In particular, the applicability of contemporary Bayesian smoothing, multiequation modeling and multivariate time series for these problems is examined. The models are large and highly parameterized, but can still be estimated effectively using Markov chain Monte Carlo. The role of Bayesian model and variable selection methodology in such models is examined, as well as the ability to obtain the full predictive distribution of key quantities of interest. While the focus is primarily on forecasting intra-day load, the case of spot prices is also examined. The models and methods are illustrated using Australian intraday data.
|
