Advancing ensemble learning techniques for residential building electricity consumption forecasting: Insight from...

Area of Science:

• Energy Systems
• Artificial Intelligence
• Sustainable Energy

Background:

• Accurate electricity consumption forecasting is vital for energy efficiency and cost management in residential buildings.
• Decision tree ensemble learning offers high accuracy for complex energy datasets.
• Explainable artificial intelligence (XAI) enhances transparency and interpretability in forecasting models.

Purpose of the Study:

• To comparatively analyze decision tree ensemble learning techniques integrated with XAI for residential energy consumption forecasting.
• To improve transparency and interpretability in short-term load forecasting models.
• To identify key influencing variables for optimized energy management.

Main Methods:

• Utilized University Residential Complex and Appliances Energy Prediction datasets.
• Applied data preprocessing and decision-tree bagging and boosting ensemble methods.
• Employed the Shapley additive explanations (SHAP) method for XAI analysis.

Main Results:

• Identified the temperature-humidity index and wind chill temperature as significant predictors for short-term load forecasting.
• Demonstrated the effectiveness of ensemble learning with XAI in explaining model decisions.
• Outperformed traditional parameters like temperature, humidity, and wind speed in forecasting accuracy.

Conclusions:

• Decision tree ensemble learning combined with XAI provides a transparent and accurate method for residential energy forecasting.
• XAI reveals non-traditional meteorological factors significantly impact energy load.
• The study promotes enhanced precision and replicability in energy system management.