Article

Good Morning,

Session 103 will be updated shortly on the UAWeek app. For your information if you plan on attending:

Title: 103: RELIABILITY-CENTERED PREDICTIVE MAINTENANCE MODELS USING MACHINE LEARNING ALGORITHMS

Description:

The Southern California Edison (SCE) Reliability Roadmap initiatives aim to enhance system reliability while promoting safety and affordability. For the Roadmap, Machine Learning (ML) based models are developed to predict asset failures, thereby reducing SAIDI/SAIFI, and increasing system reliability. This session will discuss the utilization of grid data with the inclusion of other open data sources in conjunction with ML algorithms to solve many reliability-centric problems that modern-day utilities face.

Compared to traditional, asset age driven Reliability-Centered Maintenance (RCM) methods, modern ML algorithms are capable of utilizing additional factors such as manufacture, loading, transients, weather and other environmental attributes. The models utilize different ML algorithms including Random Forests, Gradient Boosting Machines and Clustering techniques to predict asset failures. It has been proven that the SCE predictive models outperform the traditional RCM methods.

Case studies such as SCEs predictive models for the failure of transformers, switches, and poles will be presented, as well as how similar models are used to create aggregate level maintenance strategies in cables. Different implementation strategies of the models in grid operations will also be discussed.