A hybrid model of kernel density estimation and quantile regression for GEFCom2014 probabilistic load forecasting

TL;DR

This paper introduces a hybrid probabilistic load forecasting model combining kernel density estimation with quantile regression, incorporating temperature and weekly patterns, and using weighted combinations for improved accuracy.

Contribution

It presents a novel hybrid approach that integrates KDE and quantile regression with symmetry and period-based weighting for probabilistic load forecasting.

Findings

Improved forecast accuracy through temperature and weekly conditioning.

Effective combination of multiple probabilistic forecasts with period-specific weights.

Introduction of symmetry in the KDE time-decay parameter.

Abstract

We present a model for generating probabilistic forecasts by combining kernel density estimation (KDE) and quantile regression techniques, as part of the probabilistic load forecasting track of the Global Energy Forecasting Competition 2014. The KDE method is initially implemented with a time-decay parameter. We later improve this method by conditioning on the temperature or the period of the week variables to provide more accurate forecasts. Secondly, we develop a simple but effective quantile regression forecast. The novel aspects of our methodology are two-fold. First, we introduce symmetry into the time-decay parameter of the kernel density estimation based forecast. Secondly we combine three probabilistic forecasts with different weights for different periods of the month.