Stacking for Probabilistic Short-term Load Forecasting

TL;DR

This paper explores meta-learning techniques to combine various probabilistic forecasting models for short-term electricity demand, demonstrating improved accuracy through innovative local and global approaches.

Contribution

It introduces novel local and global meta-learning methods for probabilistic load forecasting, utilizing quantile regression forest and residual simulation.

Findings

Quantile regression forest outperforms other models

Local meta-learning improves forecast accuracy

Extensive tests across 35 scenarios validate effectiveness

Abstract

In this study, we delve into the realm of meta-learning to combine point base forecasts for probabilistic short-term electricity demand forecasting. Our approach encompasses the utilization of quantile linear regression, quantile regression forest, and post-processing techniques involving residual simulation to generate quantile forecasts. Furthermore, we introduce both global and local variants of meta-learning. In the local-learning mode, the meta-model is trained using patterns most similar to the query pattern.Through extensive experimental studies across 35 forecasting scenarios and employing 16 base forecasting models, our findings underscored the superiority of quantile regression forest over its competitors