Change-Point Modeling Analysis for Multi-Residential Buildings: A Case Study in South Korea

Document Type


Publication Date



Residential energy use data has become more readily available through the Advanced Metering Infrastructure (AMI). AMI data can have greater impacts in various fields that employ building energy analysis such as energy use prediction, fault detection, model calibration, and short-term monitoring because the AMI data is more meaningful due to granularity. However, this is not always true if we do not properly use the AMI data. In this paper, we evaluated how interval energy use data is useful for the prediction and short-term monitoring for residential buildings. Two phases were applied to multi-residential buildings in a case study apartment in South Korea. Phase I compares the change-point linear regression models between daily, weekly, and monthly interval energy use data. Phase II determines the minimum data period required to determine each coefficient of change-point linear regression models using an advanced analysis method compared to a previous Dry-Bulb Temperature Analysis (DBTA) study. The results from Phase I showed that weekly interval data could be the best option to analyze residential energy use. Phase II demonstrated that the new analysis method, called the coefficient checking method, is useful to find short-term energy monitoring periods for the data logger installation in terms of weather-independent and weather-dependent electricity use as well as the prediction of the whole-building energy use.