Predicting monthly electricity consumption using machine learning algorithm towards zero carbon foot print

Nassibeh Janatyan

doi:10.1007/s12667-026-00800-8

Predicting monthly electricity consumption using machine learning algorithm towards zero carbon foot print

2026-01-01

0

PRE

AI

N

Nassibeh Janatyan *

B

Boroojeni, Mohammadreza Keivani

DOI：10.1007/s12667-026-00800-8

Original

Original request for help

Save

Abstract

En 中文

Today, due to the increase of carbon emissions in the Earth's atmosphere, the Earth has experienced climate change and warming. The production and consumption of electricity have an impact on rising carbon emissions. Thus, this article aimed to lessen the quantity of carbon emissions by predicting monthly electricity consumption and offering substitute methods for its production, including clean and renewable sources other than fossil fuels like solar, biomass, and wind. This study forecasts through Decision Tree (DT) and Extreme Gradient Boosting (XG-Boost) methods for predicting monthly electricity consumption in Chaharmahal and Bakhtiari Province in Iran. This paper used six independent variables (climate, demographical changes, user type, season, year, and unexpected events) for the first time in machine learning prediction algorithms. Using GIS software and the Google Earth satellite system, by presenting the region of Chaharmahal and Bakhtiari province in Iran, they are able to acquire the quantity of precipitation, the monthly average temperature, and the amount of wind as climate factors in the algorithms. The data were then used to train and test the DT and XGBoost algorithms. Four validation metrics, i.e., Mean Absolute Error (MAE), Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), and the Coefficient of Determination (R-2), were employed to assess the algorithm's validation. The results indicate that the XGBoost method reveals fewer amounts of errors in MAE, MSE, and MAPE. The R-2 value for XGBoost (0.97184) is higher than DT (0.95345). The accuracies of these two methods (DT and XGBoost) are 0.965 and 0.996. These results show the ability of the XGBoost method to predict more accurately in this model. This model will be applied for resource management and cost planning to meet the province's electricity needs through the renewable and clean resources to lessen the carbon footprint.

Keywords:

Carbon footprint

Artificial intelligence

Decision tree (DT)

Extreme gradient boosting (XGBoost)

Weather conditions

Demographic changes

Predicting monthly electricity consumption using machine learning algorithm towards zero carbon foot print

Abstract

Journal

Organization