By Stephanie Ohnmacht, Matthew Stehman; Comcast
As the network becomes more powerful through 10G technology, resiliency in our extensive power supply network is essential to ensure our customers are always connected. The stability of outside plant (OSP) power supplies (PS) is key to keeping the network online and serving customers. Maintaining accurate location and active telemetry information is vital to keeping the power supplies in optimal health. Machine learning is employed to analyze this massive amount of telemetry data and provide actionable insights related to operating conditions and the overall health of the power supplies and batteries.
The authors, Stephanie Ohnmacht & Matt Stehman, will present a multi-tier solution that was developed to address this at scale. The solution includes the integration of mainstream mapping technology, with machine learning (ML) routines to optimize location probabilities and use of big data pipelines and advanced data science techniques to inform proactive and demand maintenance activities. Predictive models are built on top of this large dataset to help detect long term variations in power supply performance as well as real-time performance evaluations during active outages. This paper will illustrate a proven approach to improving PS resiliency, leading to increased network reliability, and satisfying the requirements of a 10G network.