San Diego Gas & Electric
Predicting the failure of utility assets
In the United States, many utilities run power lines underground in congested urban areas, or to protect them from inclement weather and wildfire risk. Yet with underground lines, equipment failures are hard to find and fix. Our utility reliability experts brought together a team of specialists to address this growing need. In partnership with industrial Internet of Things (IoT) machine learning experts Toumetis and San Diego Gas & Electric Company (SDG&E), we’re co-developing iPredict™, an innovative way to predict asset failures and we’re trialing it with the utility. SDG&E also expects to utilize iPredict™ for overhead equipment failures.
- Developed iPredict™, a first-of-its-kind asset failure predictor for electricity distribution
- Proved the technology could be a new tool for utilities and enhance electric grid reliability
- Provided the utility industry with a way to proactively plan repair work, help assure public and employee safety, and control costs
Power issues affect customers and business continuity
Undergrounding assets may increase the lifespans of power lines, as well as avoid outages caused by vehicular and tree contacts, ice and wildfires. But it also makes them harder to locate, access and repair. Crews must first locate the source of the failure, next block traffic for repairs, pump water from underground vaults and manholes into tankers, and lastly repair equipment as swiftly as possible to minimize disruption to customers. That’s why we’re co-developing iPredict™, an ingenious system that uses big data and machine learning to predict faults before they happen and reduce impact to customers.
When unplanned outages happen, utilities must make emergency repairs at an increased expense. They must also invest in resource availability to be able to respond quickly. Utilities that can predict and plan repairs and replacements can notify customers ahead of time, reduce service disruptions, and plan and execute work more cost-effectively. By giving advanced warning of disruption, utilities also enable mission-critical services, such as hospitals, fire stations and rescue services, to continue unabated.
Adopting predictive maintenance
Tee body connectors or T-splices are simple, low-cost electrical components that join mainline underground cables. But they’re currently responsible for approximately a third of the increase in The System Average Interruption Duration Index (SAIDI) at San Diego Gas & Electric Company (SDG&E). Since the company has more than 10,500 miles of underground distribution lines and an estimated 150,000 T-splices on 700 underground circuits, identifying and addressing T-splice issues is critical.
Co-developing the innovative iPredict™ solution
As an industry leader in utility reliability and the creator of the ReliabilityOne™ program, we worked with SDG&E to identify the opportunity integral in their challenges. We also provided guidance on the data required to evaluate critical assets, as well as how to use the results to maximize value. SDG&E supplied the high-frequency sub-cycle data from its electrical distribution system that we needed to populate the machine learning algorithms and provided technical subject matter expertise to facilitate running effective analysis.
iPredict™ collects and analyzes data from multiple sources, using machine learning algorithms to identify data gaps and detect potential issues with a high degree of accuracy. Utilizing high frequency electrical measurements of T-splice precursor failure signatures it predicts the impending future failures. In the future it is expected that it will be possible to tune the algorithms to identify additional types of asset failures as they show other high frequency precursor signatures. Preliminary research and testing reveal the technology could be a new tool for utilities and enhance electric grid reliability.
PA will take iPredict™ to market after completing the trials with SDG&E.