After its worst year in decades in terms of enabling consumers to keep their lights on, National Grid started talking with AiDash, an AI startup that works with utility and energy companies to maintain their satellite-powered operations.
Not only was the investor-owned gas and electric company experiencing problems with reliability, it was also struggling to contain the cost of trimming trees and other vegetation near utility lines due to increasing prices of contractors in the Northeast.
"With those contractor rates rising, we started deferring the work," said Bertram Stewart, reliability analytics and vegetation strategy manager at National Grid. "We couldn't afford to do it."
Proof of concept
After considering AiDash and other vendors that work with what the industry calls vegetation management, National Grid decided on a proof of concept with the AI startup and applied AiDash's technology to its entire Massachusetts business, which consists of about 13,500 miles of distribution line.
The vendor's AI dashboard uses satellite imagery and AI to find problems that its customers might face with different assets such as power lines, trees and circuits. The dashboard makes predictions for users about which areas need immediate attention.
National Grid opted to go with AiDash, based in San Jose, Calif., as opposed to its own data science team because the vendor was able to procure near real-time imagery of tree and shrub conditions along the utility's electrical lines. AiDash's competitors in the asset management software market include vendors Caesar and OneView.
"They don't just grab a top-down picture of our entire network, but they're actually capturing imagery based on different angles as well," Stewart said of AiDash. "It gives us a really significant degree of confidence that what they're providing to us is accurate."
National Grid's decision to apply the AiDash system to the whole distribution line in Massachusetts was a shift from its normal mode of operation. Usually, the utility company uses proof-of-concept systems on a small part of the distribution line.
The proof of concept took three months. During that time, National Grid used AiDash to identify which electric circuits to prune near based on different criteria, including how many miles of the power line had vegetation next to the wires, vegetation encroachment, data on when an individual circuit was last serviced to cut back or remove trees, and the reliability and performance history of a particular circuit.
AiDash's AI system helped National Grid understand the degree of criticality of each circuit and how badly it might perform in the next year based on how much vegetation it had around it. The technology also helped the utility company predict the probability that a circuit might malfunction or experience an outage because of overgrown vegetation.
A year of data
After about a year of using the information gained from the proof of concept, National Grid realized about $2 million in efficiencies by not having to perform as much work as it previously did, Stewart said. The company also improved its reliability performance after trending in a worse direction the previous year.
Bertram StewartManager, reliability analytics and vegetation strategy, National Grid
"We were able to reverse that performance trend and actually improve upon it," Stewart said.
Since trying out AiDash, National Grid has started to use it on a more formal basis to create future work plans. The technology enables National Grid to build work plans based on different criteria, including how much it wants to spend in the next year or which circuits it has not maintained in a few years.
"It helps us get more refined as we're doing our analysis as a result of the model that AiDash has and ultimately the output they give us," Stewart said.
A challenge National Grid ran into while using AiDash technology was getting access to the latest satellite imagery. The imagery enables National Grid to create a work plan for a circuit so that contractors can get started right away.
"We had a little delay in getting the imagery sourced, and getting the imagery processed and ingested so we can actually develop a work plan," Stewart said.
National Grid did not disclose how much it pays for AiDash's AI dashboard, but Stewart said his team's ROI time frame is short.