California’s AI-powered Wildfire Prevention Efforts Compete with Data Challenge

California-based utilities are increasingly investing in artificial intelligence to mitigate wildfires, but say they are struggling to collect enough data to train the AI.

Southern California Edison, San Diego Gas & Electric Co. and PG&E Corp.

say they see promise in AI algorithms that use images captured by drones and other means to detect infrastructure anomalies that could lead to wildfires. However, they say it will likely take years to collect enough data to deploy the algorithms at scale in their infrastructure, where they would increase ongoing manual inspections.

It’s a common problem for niche industries that don’t have easy access to massive data sets, said Gary Marcus, co-author of “Rebooting AI,” adding that AI’s effective use cases these days mostly work because we “beat them to death” with data.

“I think we’re probably a few years away from, we call it, greater penetration,” said Todd Inlander, Edison’s senior vice president of information technology and chief information officer in Southern California, adding that the AI ​​algorithms of the company are currently able to identify some specific conditions they have been trained in.

Detecting conditions from images is a two-step process, he said, in which the algorithm first learns to identify what the object is (pole, cross-arm, transformer, insulator) and then identifies whether a particular condition, such as rust or some other shape of deterioration is present.

Each of those scenarios requires hundreds or thousands of photos to train the model, he said, adding that there are “hundreds of thousands and millions of photos that we need to collect based on the different conditions.”

The company is scaling up its drone fleet to capture more photos. “For me it is a very big priority,” he said.

The consequences could be significant for these utilities. Southern California Edison said it has 188,000 miles of distribution and transmission lines. PG&E has 106,681 miles of low-voltage distribution lines and 18,466 miles of high-voltage transmission lines, many of which run through areas at high risk of wildfires, the company said.

Two years ago, PG&E agreed to use cash and stock to fund a $13.5 billion trust to compensate approximately 70,000 individuals who lost homes, businesses and family members in fires caused by the equipment.

The California Public Utilities Commission requires utilities in the state to follow a schedule of manual infrastructure inspections, but AI could complement these by making inspections more frequent and thorough, the companies said.

San Diego Gas & Electric said it has 75 working models designed to detect specific conditions or damage to assets or third-party equipment. Gabe Mika, the group’s senior product manager, said they were all trained on 100 to 5000 images. SDG&E used several machine learning and computer vision tools from Amazon Web Services to help build the models, the company said.

San Diego Gas & Electric uses images captured by drones and other means to identify problems, including cracks in infrastructure. Here is an image of equipment damage in a fire-hazardous area that has since been repaired.


Photo:

San Diego Gas & Electric

“This is one of those technologies that has such powerful, transformative potential,” said Swami Sivasubramanian, vice president of database, analytics, and machine learning for AWS, adding, “The ability to address the problem preemptively is a lot. taller.”

Mr Mika said the models were built primarily to identify conditions from aerial footage captured by drones, but recently the company has begun investing in cameras on fleet vehicles to capture more footage. He said the first models may require additional training to identify the conditions as captured by vehicle cameras, taken from different angles.

According to Dr. Marcus cannot generalize or reason from small amounts of information today’s AI algorithms like humans can. They make up for it by taking in massive amounts of data. That said, he added, the more the incoming data resembles the data the algorithm was originally trained on to, the less you’ll need.

dr. Marcus is also the founder of Robust, a software development startup. AI.

PG&E says it has had some success with computer vision models designed to find things like insulator fouling with each model trained on thousands to tens of thousands of images. Smaller anomalies such as cracks are harder for models to spot because they can be masked by dirt, guano or water, said Andy Abranches, PG&E’s senior director of wildfire risk management.

The goal, he said, is to prioritize building models that recognize anomalies that could lead to equipment failure and cause ignitions. To do that, when there is a breakdown, the company has to go through years of previous images of that device to build a model that recognizes the signs.

“The depth of data you need will probably take a few more years.” [collecting] data sets,” said Mr Abranches.

write to Isabelle Bousquette at [email protected]

Copyright ©2022 Dow Jones & Company, Inc. All rights reserved. 87990cbe856818d5eddac44c7b1cdeb8

Leave a Comment