Anomaly detection for proactive equipment maintenance and energy management at supermarkets

Lead University: Carnegie Mellon University
PI: Burcu Akinci, Civil and Environmental Engineering
Co-PIs: Mario Berges, Civil and Environmental Engineering

The current practice of maintenance in supermarkets is mainly reactive, in which maintenance happens only after equipment failure or inefficiency occurs. This, however, can bring a significant reduction in the service level as well as energy performance. Therefore, it is needed to automatically detect such anomaly and failure in advance so that facility managers can take actions proactively and strategically before substantial loss occurs.

In this research, we propose to develop a method for automatically detecting anomalies from electricity consumption data of a grocery store to predict malfunctioning, failure, and inefficient operation of lighting, refrigerators and other major load categories. We focus on lighting and refrigerators because those are the major electricity consumer in supermarkets (80 percent of electricity in a mid-sized supermarket) and also closely related to the customer’s service satisfaction. By applying advanced statistical anomaly detection techniques to the energy consumption record and equipment maintenance log in a grocery store, we will find a model that maps energy consumption and equipment replacement.

In partnership with Giant Eagle, we will have access to an extensive historical record of both energy usage patterns.