Day-ahead electricity price forecasting with high-dimensional structures: Univariate vs. multivariate models

Conducting an extensive empirical study on short-term electricity price forecasting (EPF), involving state-of-the-art parsimonious expert models as benchmarks, datasets from 12 power markets and 32 multi-parameter regression models estimated via the lasso, we show that using the latter shrinkage approach can bring statistically significant accuracy gains compared to commonly-used EPF models. We also address the long-standing question on the optimal model structure for EPF. We provide evidence that despite a minor edge in predictive performance overall, the multivariate modeling approach does not uniformly outperform univariate models across all datasets, seasons of the year or hours of the day, and at times is outperformed by the latter. This may be an indication that combining advanced structures or the corresponding forecasts from both modeling classes may bring a further improvement in forecasting accuracy. Finally, we also analyze variable selection for the best performing multivariate and univariate high-dimensional lasso-type models, thus provide guidelines to structuring better performing forecasting model designs.