Online Program

Return to main conference page
Friday, October 4
Fri, Oct 4, 2:30 PM - 3:45 PM
Evergreen I
Speed Session 4

Battery Wear Modeling for Predictive Maintenance (306530)

*Anna Lake, Schneider Electric 

Keywords: IoT, Data Center, Batteries, Predictive Maintenance

For a Data Center Operator/Manager, constant uptime of IT equipment (e.g. servers) is the highest priority. To ensure this, they rely on uninterruptible power supplies (UPSs) to provide battery backup power in the case of an outage. The health of a UPS battery supply is critically important to the function of the device, mandating service contracts to regularly monitor battery health. Service visits are typically scheduled based on time (e.g. once a year) and are not always necessary. With the use of engineering-defined battery wear models, real-time IoT sensor data on environmental and usage parameters, and pilot testing for model feedback, we are able to remotely estimate the current battery state of health (%) and approach service / replacement scheduling in a more intelligent and dynamic manner by forecasting when UPS batteries will reach “end-of-life” (EOL). This project focuses on accurately estimating battery wear by calibrating model components to varying manufacturer and chemical specifications as well as varying sensor characteristics.