Backup batteries are an essential part of the reliability equation for the cable outside plant architecture. When utility power fails, batteries are relied upon to power the network and helps ensure customers remain on-line. Being able to predict how long a battery can support the network is an essential planning and maintenance tool.
As one of the largest industrial Internet-of-Things (IOT) applications, Comcast has developed the unique ability to automatically test and monitor over a million installed batteries at well over a quarter of a million installations. This paper describes possible ways to use this continuous sensing of metrics from the power supplies of our outside plant network to begin developing a sophisticated diagnostic and planning tool. When finalized, this tool will use machine learning and artificial intelligence to be able to predict the expected runtime of a battery during a utility outage based on its actual load, and then to adjust this runtime prediction based on multiple key factors. The final desired result is the ability to know whether a battery is performing as expected, and to be able to track its degradation for preemptive maintenance purposes. This includes unexpected loss of battery performance from unknown factors as well as the expected losses from known factors. The status of this work is presented, showing the current predictive model, along with a brief discussion of the future work planned in this area.