Event Detection Using Linear Regression and Historical Weather Data

Abstract

The Research Triangle Institute (“RTI”) seeks to minimize water consumption by automating the process of detecting water and energy over- and under-consumption events associated with cooling systems. The Central Utility Plant (“CUP”), which serves roughly 25% of gross built area on RTI’s main campus, has contributed to past over-consumption events due to mechanical failure of cooling tower water makeup float valves. RTI’s facilities team would like to assemble data and examine the relationship between atmospheric conditions and water consumption. This project entails development of a data cleaning and analysis tool based in Microsoft Excel that allows RTI facilities and operations teams to periodically update a predictive model in response to changing facility parameters that are external to the model, including changes in building footprint, occupancy and HVAC settings. The final deliverable includes a user guide that explains the functions of the Excel tool as well as the limitations of predictions based on linear regression models.