A Machine Learning-Based Reference Baseline Consumption Model for Energy Efficiency Measurement and Verification

In Energy Efficiency Efficiency Projects, the IPMVP (International Performance Measurement and Verification Protocol) methodologies are employed to measure and verify energy savings based on pre- and post-intervention analyses of energy consumption However, these methods, which depend on low-resolution data and basic regression models, often fall short in accurately calculating target reference consumption and efficiency. Advances in measurement technology now enable real-time energy efficiency analyses with high-resolution data. Additionally, integrating these datasets with sophisticated machine learning models has been observed to improve the accuracy of reference consumption calculations. This study aims to improve the consistency of reference consumption calculations by developing machine learning-based reference consumption models. The piecewise regression, SARIMA, and XGBoost models developed in this research were tested across four distinct facilities equipped with high-resolution data, and their performance was systematically evaluated.