EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0843
Title: Forecasting half-hourly electricity prices using a mixed-frequency VAR framework: The case of New Zealand market Authors:  Nuttanan Wichitaksorn - Auckland University of Technology (New Zealand) [presenting]
Gaurav Kapoor - Auckland University of Technology (New Zealand)
Wenjun Zhang - Auckland University of Technology (New Zealand)
Abstract: A mixed-frequency vector autoregressive (VAR) framework is employed to forecast half-hourly electricity prices while having several weather variables and electricity demand that come with another frequency. In addition to a standard VAR model used in the analysis, we propose a VAR model extending from a single-equation RU-MIDAS model. LASSO is also incorporated to help with the variable selection. These models are estimated using a range of techniques including least squares, Gibbs sampling, and Variational Bayes. We compare our forecasting results with those from random subspace regressions, e.g., subset and projection regressions. We found our results are favorable, especially those with LASSO.