Enhancing residential load forecasting accuracy through dynamic feature selection and ensemble machine learning models: A real-world scenario in Southern Finland - UTU Research Portal

A1 Refereed original research article in a scientific journal

Enhancing residential load forecasting accuracy through dynamic feature selection and ensemble machine learning models: A real-world scenario in Southern Finland

Authors: Taheri, Nabi; Karttunen, Lauri; Jouttijärvi, Sami; Piazzi, Antonio; Tucci, Mauro; Miettunen, Kati

Publisher: ELSEVIER SCIENCE SA

Publication year: 2025

Journal: Energy and Buildings

Article number: 116589

Volume: 349

ISSN: 0378-7788

eISSN: 1872-6178

DOI: https://doi.org/10.1016/j.enbuild.2025.116589

Publication's open availability at the time of reporting: Open Access

Publication channel's open availability : Partially Open Access publication channel

Web address : https://doi.org/10.1016/j.enbuild.2025.116589

Self-archived copy’s web address: https://research.utu.fi/converis/portal/detail/Publication/505439929

Abstract

This study aims to improve the accuracy of residential electricity consumption forecasts, a key component of modern energy management systems, especially in settings with high demand variability. The novelty of this work lies in combining dynamic feature selection-adapting to recent data patterns-with a stacking-based ensemble model that leverages multiple predictors. A comprehensive hourly dataset from 200 residential buildings in southwestern Finland was collected throughout 2023, including meteorological variables and real-time electricity prices. The forecasting task was defined as a day-ahead (24-hour ahead) prediction. The year was divided into cold (October-March) and warm (April-September) periods. Four machine learning models were employed per season, including XGBoost, Random Forest, Voting, and a Stacking ensemble with Ridge regression as the meta-learner. The stacking model achieved the best performance in 163 buildings in the cold period and 178 buildings in the warm period. Feature importance was assessed using SHAP values, comparing static and dynamic feature selection strategies. The dynamic approach reduced average prediction error from 11.85% to 9.31% in cold months and from 11.67% to 9.14% in warm months, outperforming the static method in over 96% of buildings. These findings underscore the effectiveness of adaptive feature selection and ensemble learning in capturing seasonal and behavioral dynamics in residential electricity usage.

Downloadable publication

This is an electronic reprint of the original article.
This reprint may differ from the original in pagination and typographic detail. Please cite the original version.

1-s2.0-S0378778825013192-main.pdf

Funding information in the publication:
Nabi Taheri gratefully acknowledges the support from the University of Pisa, the University of Turku, and i-EM Company during the preparation of this work. This project was funded by the Strategic Research Council (SRC) established within the Research Council of Finland (RealSolar, Decision No. 358542). The work was supported by the Strategic Research Council within the Research Council of Finland, Decision No. 358542 (KM, SJ). Lauri Karttunen thanks the City of Salo and the University of Turku (project HEMS), and UTUGS.