Tapestry, the Google X project tackling the electric grid, has so far focused its public partnerships on bringing artificial intelligence tools to transmission challenges. In Chile, the national grid operator is using Tapestry’s tools for its annual transmission planning process. And North America’s largest grid operator, PJM, is applying Tapestry’s AI tools to its massive interconnection backlog.
But as the project’s transmission efforts unfold, Tapestry has also been quietly working on tools for the distribution grid. Today, Latitude Media has learned, Tapestry is unveiling a key milestone in that work: a partnership with a distribution utility in New Zealand.
Vector, the largest of New Zealand’s 29 distribution utilities, is now using Tapestry’s grid management and planning tools across its network for daily operations — the technology’s first scaled deployment on a distribution network. Grid Aware, Tapestry’s AI-powered asset inspection tool, has already cut Vector’s average inspection times from 45 minutes to around five minutes per asset. That additional speed and accuracy gave Vector the network visibility it needed to also deploy Tapestry’s grid planning tool, which simulates future scenarios to plan for resilience and reliability.
Tapestry’s transmission tools actually are the result of the project’s distribution-focused work, explained Page Crahan, Tapestry’s general manager. When Tapestry was first created inside X, the plan was to tackle distribution-level challenges first.
Distribution grids are “less understood, less measured, less mapped with high confidence than the transmission network,” Crahan told Latitude Media. “One of the things that’s really challenging for distribution network operators is getting a high confidence…representation of their current network from which they can make decisions.”
But in Tapestry’s early years — while the team was “making progress on distribution tools” but didn’t yet have them deployment-ready — the global energy landscape started to change, and fast.
First and perhaps most importantly, the narrative around load growth shifted. In 2018, the focus was on load growth from crypto or electric vehicles. In a matter of years, though, load growth from artificial intelligence and widespread industrial electrification was both larger and more urgent. Meanwhile, advancements in AI and machine learning were changing Tapestry’s work as well, Crahan said, and significantly boosting the team’s capabilities.
“Tapestry looked at that trend, probably a little bit early…and we knew that there was an ‘all hands on deck’ [moment] for transmission planning coming immediately,” she explained. “When I think about managing resources on our team and where we should focus, it wasn’t like distribution was solved so we should stop and put our pencils down…it was more about doubling down on things that felt really urgent at the time.”
Working around the data problem
When Crahan and her team first met with Vector in 2019, the utility was touring innovation hubs around North America. The utility was in search of tools to ensure they’d be ready if and when Auckland’s 1.7 million residents decided to electrify their homes and lives.
But the COVID-19 pandemic hit a few short months after those initial meetings, changing how the teams could work together.
“Things went a little more slowly at the beginning,” Crahan said. The upshot of that, she added, was the Tapestry team “really understanding the problem before we started building things, because it was the best we could do remotely.”
For example, Tapestry and Vector initially set out to build a distribution planning tool. As they worked together throughout the slower months of the pandemic, they realized Vector first needed a better understanding of the real-time health of the network — before planning for future expansion. That need led to the creation of the Grid Aware tool. The automation of inspections and defect detection are great for preventative maintenance, Crahan said, but more importantly they provide “information that was foundational to driving the planning tool.”
Grid Aware enables partners to pull together images of their assets (from utility poles to transformers) from multiple sources, including satellite and street-view imagery, as well as images taken by field teams. (In the case of Vector, this includes the helicopter and drone images the utility was already collecting as well.) The tool combines those various forms of visual data into a single view to simplify inspections.
The key to training the model, Crahan said, was expert annotation. Experienced Vector field crews labeled visual features in the images, essentially teaching the system to recognize important cues. This human-in-the-loop approach was key, she added: “Ask someone who’s spent their entire career evaluating and maintaining a network, they can look in less than 15 seconds and see things that would be incredibly challenging to train a machine learning model to do.”
The distribution impact
With that foundational data in place, the pair turned to Tapestry’s grid planning tool. But it wasn’t just a matter of copying over the work Tapestry had done on transmission grids. A distribution tool needed custom models, interfaces, and workflows designed to address the operational realities and decision-making needs of distribution utilities, Crahan explained.
Transmission and distribution planning do share key steps, she added, like preparing future scenarios, running power flow and economic simulations, and analyzing system constraints, but the networks are fundamentally different.
A transmission planning tool like the one Tapestry has deployed in Chile and PJM simulates scenarios for high-voltage, long-distance power flows, and considers large-scale expansions like load growth from industry or population centers. A distribution planning tool needed to consider a wider range of localized contingencies and operational constraints. Those are often managed by many separate tools, which makes scenario building and modeling more complex, Crahan said.
Now though, seven years on from Tapestry’s inception, the deployment of their tools on a distribution grid provides a glimpse of the project’s broader strategy finally coming to fruition. With Vector, Crahan said, Tapestry is “beginning to connect transmission and distribution tools and simulations to create efficient singular solutioning.”
From an even broader perspective, though, Tapestry sees the deployment as a demonstration of how AI can help the industry meet energy demand, rather than just create it.
“This is another story of where machine learning can be used — not in a lab, not in a hypothetical,” Crahan said. “A field worker is going to do their job faster and more easily this afternoon in Auckland because of this machine learning work.”
