California’s wildfires are 500% larger due to climate change, burning 4.3MM acres in 2020. Increasing the risk of wildfires reaching the power grid and the risks of ignition. Electric utility SCE serves 50,000 sq miles (area equaling Greece), with 14,000 miles of power lines and 300,000 poles in high fire risk areas. Utility line workers relied on ground inspection to monitor risks posed to the grid. Due to the rapid acceleration of climate change, it was becoming increasingly challenging to monitor changing conditions of the grid by foot from below.

Our line workers were previously capturing field inspection images (pole deterioration, cross arm damage, and vegetation overgrowth) and manually uploading to an app for risk review. Far too often, these images were unusable, not machine readable, and less actionable for maintenance and remediation. SCE was determined to capture high quality data inputs for improvements in decision making. To tackle wildfire risk, SCE had to see better, faster, and at scale.

Each inspection requires repeated data inputs. Thousands of inspection images were already collected daily. Existing workflow could become the hinge for systemic improvement. The first challenge was getting better pictures. An intuitive human-centered AR app (Inspect Cam) was built to open within SCE’s existing ground inspection app after extensive usability research. Inspect Cam opened as Assisted Reality overlay on a new screen, providing guidance on framing, aspect, orientation, and other standard inputs to improve image captures. Inspect Cam’s improved image data were the first steps in a transformation that combined computer vision and AI to enable a partially automated inspection process.

The next phase invented and prototyped a custom Computer Vision AI layer, partially aiming to automate simple questions within the inspections process by prefilling answers using AI based on the high quality images captured: which assets were present (poles, insulators etc.), how many assets on the pole, and others SCE determined could be good candidates for automation. Coupling AI data analysis and Assisted Reality, SCE captured high quality images together with grid asset data. These new Computer Vision tools increased both accuracy and speed in the inspections process and improved overall image readability and usability across the organization.

With the app in full use by ground inspection teams, the prototyping phase was now complete and overall visual data quality dramatically improved, with machine legibility increasing from 32% to 78% using AR. 33% of inspection questions were automated, which reduced inspection cycle time. With this experience proving successful in the field, we expanded our computer vision program. Now, we’re training ground inspectors to fly drones to get a 360 view of thousands of poles from above – directly from their tablets. We implemented machine learning to scan thousands of live images per day, to detect hazards like dry vegetation near power lines. And we built a proprietary AI to anticipate risks like pole deterioration and cross arm damage. In 2021, the system helped analyze 200,000 poles, and found 800 at critical risk of ignition, safely sending experts to fix them in less than 24 hours. The strategic impacts enabled better maintenance decision-making from improved quality of field inspections. More accurate and targeted remediation decisions can be made.

These operational and process changes are transforming the way we monitor, maintain, and repair the distribution system – turning our power grid into a smart wildfire mitigation system.

In 2021, The Prevention Grid helped analyze 200,000 poles, and found 800 at critical risk of ignition, safely sending experts to fix them in less than 24 hours.

With grid-hardening work completed last year, The Prevention Grid is helping to reduce power shutoff outage time by an estimated 70% on frequently impacted circuits (based on 2021 weather and fuel conditions).

Lowering the risk of catastrophic wildfire damage by 65% (relative to pre-2018 levels).

Helping reduce carbon emissions caused by wildfires hitting the power grid – going from 128,000 acres burned in 2020 to less than 500 in 2021.

The Prevention Grid has strengthened grid security, dramatically cut multi-attribute risk scores, and helped keep the power on for millions. Ultimately, transforming our power company into one of the largest aerial remote sensing programs in the US.