Analysis of Variance Combined with Optimized Gradient Boosting Machines for Enhanced Load Recognition in Home Energy...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Energy Systems

Background:

• Current Home Energy Management Systems (HEMSs) have limitations in comprehensive load recognition.
• Existing methods struggle with accurate appliance identification and robust model performance.
• There is a need for improved feature selection and inter-class separability in load recognition.

Purpose of the Study:

• To propose a novel load recognition approach for HEMSs.
• To enhance appliance identification and system performance using advanced machine learning techniques.
• To improve feature selection and inter-class separability for more reliable load recognition.

Main Methods:

• Utilized Analysis of Variance (ANOVA) F-test for feature selection.
• Combined ANOVA with SelectKBest and gradient-boosting machines (GBMs) including HistGBM, LightGBM, and XGBoost.
• Optimized GBM models to improve the reliability of the load-recognition system.

Main Results:

• The ANOVA-GBM approach demonstrated superior training time efficiency compared to Principal Component Analysis (PCA).
• ANOVA-XGBoost, ANOVA-LightGBM, and ANOVA-HistGBM were significantly faster than their PCA counterparts.
• Achieved high accuracy (up to 96.75%), F1 scores (up to 96.64%), and Kappa indices (up to 0.9452), surpassing existing methods.

Conclusions:

• The proposed ANOVA-based feature selection combined with GBMs offers a more effective and refined load recognition system for HEMSs.
• The approach significantly improves accuracy and efficiency, addressing limitations in current load recognition techniques.
• The method provides substantial accuracy gains, highlighting its potential for practical HEMS applications.