Measurement and Verification Building Energy Prediction (MVBEP): An interpretable data-driven model development and analysis framework.

Due to the rapid development of Advanced Metering Infrastructure (AMI), data-driven approaches are becoming more effective and accurate for Measurement and Verification (M&V) baseline energy modeling of existing buildings using historical energy consumption data. This paper develops and applies a comprehensive analysis framework suitable for developing baseline data-driven models to estimate energy savings from building retrofits. The framework employs common data-driven modeling approaches for building energy prediction, necessary data processing tasks, and industry-used evaluation methods. The modeling approaches considered by the developed framework include linear regression, ensemble models, support vector regression, neural networks, and kernel regression. The developed framework is applied to two case studies with different sources of energy consumption data generated via a simulated office building model and actual measurements obtained from a publicly available dataset. The framework applications to data from 170 buildings with 1-year data result in energy use predictions with medians of Goodness-of-Fit (GOF) scores that are less than 25%. • A data-driven framework is proposed to automate baseline modeling for measurement and verification analysis. • The framework considers several data-driven modeling approaches and timestamp frequencies. • The framework is able to predict within 5.54%, the energy savings incurred by retrofitting an office building. [ABSTRACT FROM AUTHOR]