Machine Learning-Driven Urban Energy Consumption Prediction in Norway: A Case Study of Oslo

Abstract

Urban energy consumption prediction is critical for Norway’s transition to a low-carbon society, as cities account for 75% of the country’s total energy use. This study applies four machine learning models—XGBoost, LightGBM, CatBoost, and Multilayer Perceptron (MLP)—to predict short-term (24-hour) urban energy consumption in Oslo, using a comprehensive dataset (2022–2023) including meteorological data, building characteristics, and historical consumption records. Experimental results show that LightGBM outperforms other models: it achieves a Mean Absolute Error (MAE) of 0.32 GWh, Root Mean Squared Error (RMSE) of 0.45 GWh, and Mean Absolute Percentage Error (MAPE) of 4.12%, which are 15.8%, 18.2%, and 16.5% lower than XGBoost, CatBoost, and MLP, respectively. The model’s ability to capture non-linear relationships between temperature fluctuations (a key factor in Norway’s cold climate) and energy use provides valuable insights for Oslo’s energy grid management and renewable energy integration.

Previous article in issue Next article in issue