Register to attend the PG&E Innovation Summit presented by DISTRIBUTECH. The event, taking place on Nov. 13 in San Jose, California, is free to attend but requires advanced registration and approval. More information here.
As the holiday season approaches, children worldwide are drafting and re-drafting their pleas to a certain toy kingpin up north. Their requests are candid and detailed to avoid confusion and play an outsized role in ensuring a joyful yuletide.
But for a utility in pursuit of lowering costs for customers and fighting the impacts of climate change, Santa Claus, venerable as he may be, is of little use. Pacific Gas & Electric Company has penned its own wish list of sorts, this one addressed to the technology innovators capable of helping the utility “unleash the power of AI.”
Ahead of the upcoming PG&E Innovation Summit presented by DISTRIBUTECH, PG&E released its second annual research and development strategy report— 195 pages outlining the challenges and opportunities of creating an “AI-enabled utility.” While utilities are criticized, often fairly, for purporting a Do No Wrong narrative, PG&E is choosing to magnify its gaps and call on the market to fill them.
The Innovation Summit, taking place on Nov. 13 in San Jose, California, will probe the 67 high-priority challenges facing PG&E’s strategy while highlighting the utility’s AI initiatives already underway.
“At PG&E, we are actively building the carbon-free energy system of the future,” PG&E CEO Patti Poppe said. “A system in which advanced automation refines our operations, predictive technologies prevent equipment failure, and personalized experiences benefit every customer.”
How PG&E is using AI today
Through a pragmatic approach to scaling AI tools, PG&E said it’s ready to “move beyond pilot developments” to implement a “holistic” strategy. Several initiatives are already underway, addressing power grid asset health, wildfire mitigation, and customer engagement.
In what PG&E teams dubbed “Sherlock & Waldo,” the utility’s system inspection group developed a suite of tools that enable desktop inspectors to efficiently view and mark potential equipment problems on high-resolution images captured through aerial inspections using drones and helicopters. The images train machine learning models to identify and classify electric system components. PG&E said the tools have saved 7,500 inspector hours annually and provided over 600,000 grid inventory data entries to support strategic maintenance and management of its electric assets. Additional machine learning models allow PG&E to assess the probability of whether a transformer will fail within the next 30 days and flag high-risk transformers for desktop review.
PG&E built machine learning models that can process real-time weather data to proactively assess wildfire risk over a forward-looking 129-hour window or approximately 5.5 days of enhanced visibility. This look-ahead guides how PG&E operates the grid in targeted locations based on the predicted timing and location of elevated fire danger, further advancing our ability to eliminate ignitions.
For improved and efficient customer engagement, PG&E is utilizing intelligent conversational assistants and AI-powered call center software that can route calls more efficiently and personalize service with real-time, relevant information specific to the customer.
What comes next
Despite PG&E’s optimism around AI, and desire to emerge from the all-too-familiar pilot loop that plagues utilities, they remain “conservative” in approaching new tools. AI tools are at risk for hostile attacks, model misuse, and model failure, the utility said, while some tools — like generative AI — are still too immature to introduce to some aspects of the system.
PG&E added 17 new problem statements since last year’s R&D report. Each is categorized by business segment — electric, gas, system-wide — and assessed for low, medium, or high AI potential. Among the additions are increasing the precision and efficacy of energy management programs (high AI potential); accelerating new interconnection planning (high AI potential); and optimizing soils and spoils management for underground construction (low AI potential).
GO DEEPER: Can a utility lead the energy transition? Inside the rebirth of PG&E
Each chapter of PG&E’s R&D report opens with a vision for the future and closes with a pointed request. On electric vehicles, PG&E asked vendors for solutions to “reduce the costs of supporting V2X capabilities, improve data visibility, and create new avenues to maximize the potential of EVs as grid assets.”
“We sought to embrace breakthrough thinking and radical collaboration by openly sharing our toughest technological challenges and inviting problem solvers from around the world to help PG&E envision what the future could look like,” said Quinn Nakayama, PG&E’s senior director for grid research, innovation, and development. “We look forward to partnering with you—innovators, regulators, educators, and peers—to co-create the solutions that will unlock a future that is cleaner, safer, more affordable, and more reliable for all Californians.”
No one gets everything on their Christmas list, but PG&E’s forthcoming approach seems to be working. The utility retired 17 problem statements from the 2023 R&D report and reported minor or major updates to 35.
The future, Poppi said, is “ours to imagine and create, and together we will achieve extraordinary things.”
