Integrating Causal Analysis with Machine Learning for Meteorological-Based Energy Consumption Forecasting: A Multi-Method Approach
International Journal of Informatics and Computation · 31.07.2025 · 7 (2) · 384-399
DOI: 10.35842/ijicom.v7i2.134 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Informatics and Computation, online erschienen am 31.07.2025. Verfasst von Zikri Wahyuzi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
In the era of climate-aware infrastructure planning, accurate and interpretable energy consumption forecasting is a critical need for policymakers and utility providers. Traditional forecasting models often struggle to capture non-linear and dynamic relationships between meteorological factors and energy use. In this study, we propose a hybrid framework that integrates causal analysis and machine learning to enhance both the accuracy and interpretability of energy consumption prediction. We applied three causal inference techniques, including Granger causality, partial correlation, and mutual information, to identify the most influential weather variables affecting energy use from a dataset of 26,323 hourly records. The results showed that temperature-related variables, particularly dry bulb temperature (MI = 0.0529), were the most causally significant predictors. We then trained five machine learning models: Random Forest, Gradient Boosting, Decision Tree, Ridge Regression, and a Multi-Layer Perceptron Neural Network. Among these, Random Forest achieved the best performance with MAE = 0.0501. Additionally, we performed a temporal correlation analysis showing significant hourly shifts in temperature-energy relationships, offering insights for time-sensitive energy demand strategies. This framework bridges the gap between predictive accuracy and causal understanding, making it valuable for smart grid management and climate-resilient energy planning. It also offers practical adaptability for diverse geographic settings and evolving energy infrastructures.
Zitieren
APA
Wahyuzi, Z., & Abdurrahim, A. (2025). Integrating Causal Analysis with Machine Learning for Meteorological-Based Energy Consumption Forecasting: A Multi-Method Approach. International Journal of Informatics and Computation, 7(2), 384-399. https://doi.org/10.35842/ijicom.v7i2.134
Vancouver
Wahyuzi Z, Abdurrahim A. Integrating Causal Analysis with Machine Learning for Meteorological-Based Energy Consumption Forecasting: A Multi-Method Approach. International Journal of Informatics and Computation. 2025;7(2):384-399. doi:10.35842/ijicom.v7i2.134
BibTeX
@article{Wahyuzi2025,
author = {Wahyuzi, Zikri and Abdurrahim, Abdurrahim},
title = {Integrating Causal Analysis with Machine Learning for Meteorological-Based Energy Consumption Forecasting: A Multi-Method Approach},
journal = {International Journal of Informatics and Computation},
year = {2025},
volume = {7},
number = {2},
pages = {384--399},
doi = {10.35842/ijicom.v7i2.134}
}
EUVIMED