Incipient Failure Identification for Common Grid Asset Classes

Project Description

Individual sensing of all components on the grid is a challenge, the grid is aging, and the approaches to identifying critical failures on so many components are limited by data silos and specific analytics for each componentThe project will, in three years, operationalize a multi variate, multi modal approach to diagnose and prescribe remediation pathways for both short term but critical failures locally and incipient growing problems centrally in commonly utilized equipment throughout the country.

Value Proposition

  • Incipient Failure of devices can impact resiliency, reliability and safety—in particular in regions with high fire and natural impact issues. Mitigation requires multi-variate, multi modal approaches to address local safety, locate and isolate intermittent failures, and better forecast maintenance needs.
  • This project derives directly from gaps identified in the application of machine learning to grid modernization to solve critical resilience, reliability and safety problems in the distribution field. These techniques can be applied to any grid assets with multi-modal data collection. Use cases are highly visible ”real” problems, which pure sensing can't fix.
  • Alternate approaches require massive investments—for example insulating all overhead lines in California would cost billions of dollars.

Project Objectives

Develop and test a multi-stage approach for identifying failures, dealing with local sensing and “on the ground” analytics, improvement in detection of growing failures, and addressing diagnostic and prognostic indicators.  The project will integrate state-of-the-art sensing with key outcomes from prior GMLC projects in signature identification, sensor development and machine learning, in order to rapidly identify common causes of both reliability, and resiliency incidents related to devices.