As Oregon’s largest utilities seek to better protect themselves and their customers from the impacts of wildfires and other extreme weather events caused by climate change, they are increasingly harnessing the power of big data analytics and artificial intelligence to guide them.
Both PacifiCorp and Portland General Electric are plowing millions of dollars into systems to forecast weather and fire risks at a much more granular level across their service territories, track their transmission and distribution systems in real time, and model potential outages and risks during a given weather event.
It’s a fast-evolving approach to the way utilities prepare for and respond to extreme weather, one designed to guide their investments in tree-trimming and upgrades to transmission and distribution systems, improve their situational awareness, and provide predictive capabilities on where and when to pre-position crews or turn off the power during periods of high wildfire risk.
The initial catalyst for Oregon utilities to bolster their wildfire preparedness was 2018′s Camp fire in California, where a faulty electric line ignited a firestorm that burned 18,000 structures in the town of Paradise and caused at least 85 fatalities. California’s largest electric utility, Pacific Gas and Electric, filed for bankruptcy the following year, citing expected wildfire liabilities of $30 billion.
Those efforts became far more urgent in the wake of Oregon’s devastating Labor Day fires in 2020. PacifiCorp is facing a raft of lawsuits alleging that its equipment was responsible for igniting several of those conflagrations. PGE, by contrast, instituted its first ever public safety power shutoff in the Mount Hood corridor. It avoided any problems in that area, but still saw a big swath of customers southeast of Portland impacted by the Riverside fire.
The fires pushed the Legislature to pass Senate Bill 762, a sweeping 2021 wildfire preparedness and mitigation bill. Among other provisions, it requires utilities to have their wildfire preparedness plans approved by the Public Utility Commission, including their plans for preemptive blackouts. A foundational piece of that preparedness for both utilities is leveraging new technology to more accurately assess risks across their systems.
It’s a lot of ratepayer money, a lot of computing power and a whole lot of data. But acting on the insights could save lives, buildings, costly equipment and avoid potentially massive liabilities from utility-caused fires.
Letha Tawney, one of Oregon’s three public utility commissioners, has been the agency’s point person on wildfire preparedness. She says it’s difficult to peel apart how effective such technologies have been when implemented by utilities in California in an age when climate change is driving such extreme weather. But she said the commission will be part of that conversation as utilities seek to raise rates to recover the significant investments they are making.
“It’s all very new. The question of performance and how to know if something is effective is a very live conversation in the regulatory community,” she said. “It’s a dialogue we’ll have to be in. Success will be when we can say, ‘That is effective and that is less effective and here’s how we’re changing it.’”
From the utilities’ perspective, they have little choice. Whether it’s an ice storm, a spring snow storm, extreme rains, a heat dome or wildfires, “we’re looking at extreme weather happening more frequently and impacting the electric system,” said Allen Berreth, PacifiCorp’s vice president of transmission and distribution operations.
In response, the company plans to invest $500 million in wildfire mitigation across its six-state territory by 2028, he said. That spending will take years to play out, and most of it will go to increasing the resiliency of its electric lines by, for instance, installing covered conductors, ductile iron poles or burying some circuits underground. In the meantime, it is leaning heavily on data analytics to guide those investment decisions, predict the weather and analyze risk.
PacifiCorp used to outsource much of that analytical and forecasting work. But in February 2021, two days before the big ice storm that left hundreds of thousands of Oregonians without power, the utility brought on Steve Vanderburg as manager of its new in-house meteorology team. Vanderburg started his career at the National Weather Service, where he managed the red flag warning program in the San Diego area. He spent the next decade at San Diego Gas and Electric building a wildfire mitigation program in response to increasingly extreme fire weather experienced in that region.
As those condition have migrated north, that work by California utilities has now become a blueprint for what’s taking place here, Vanderburg said. He spent the last 16 months building out a high-resolution forecasting model that goes out 96 hours and tells the utility not only what the temperature, humidity and winds are expected to do, but provides very localized fire weather indices such as the moisture level in live and dead vegetation, the expected rate of spread of any ignition and how they are predicted to change over time.
That’s pretty much how meteorologists have forecast weather and fire risks for decades, said Vanderburg, but the technology and computing horsepower keep getting better. The model will eventually incorporate hourly weather and the company’s outage history over the last 30 years. That will allow it to compare a given forecast to every day for the last three decades, predict how extreme conditions will be and the likely impact, whether it’s an outage on a specific circuit or a wildfire ignition and how that fire will behave.
“It allows you to focus in on what’s important,” Vanderburg said. “This is what allows us to start answering those questions days in advance, of where do we look, what are the potential impacts and how are we going to mitigate this risk.”
The company is also delivering its forecast data to an outside vendor, Techno Silva, that models how wildfires are likely to expand in specific locales. The model simulates fire conditions over the forecast period, allowing the utility to evaluate fire size potential in specific areas, buildings and population impacted, rates of spread, flame length and identify circuits at highest risk.
That information is highly location-specific, as 35 mile-per-hour winds, for example, could lead to very different consequences in different locations based on vegetation, drought conditions and the age of local equipment.
If the utility knows there are high weather-related outage risks or high wildfire risks, “there are things we can do ahead of time to prepare our system for that risk,” Vanderburg said. “That’s how we’re going to use these tools to stay ahead of the weather.”
To increase real time situational awareness, the utility is expanding its network of weather stations across its service territory, which includes more than 600,000 customers scattered in smaller cities and rural areas throughout the state. It is also pulling data from stations managed by the National Weather System and agencies such as the Oregon Department of Transportation, providing some analytical tools and making them available to the public at pacificpowerweather.com.
Likewise, PGE is planning to invest $110 million over the next five years on technologies, data platforms and expertise to enhance the resiliency of its system and improve situational awareness across the 900,000 customers it serves in a territory that stretches from the Columbia River to south of Salem, and the Mount Hood area to Pacific City on the coast.
A team of data scientists, mapping specialists and meteorologists began working last year with an outside vendor, mPrest, to build a data analytics platform that provides a real-time visualization of its distribution system and assesses risks to it based on current and forecasted weather conditions.
The model is populated with hundreds of different variables, including detailed information on all of PGE’s assets; their age and outage history; local vegetation; weather; fuel and soil conditions; homes and structures in the area; the probability of ignition detection; and likely response times in specific areas. It provides a risk score to guide decisions on whether to implement public safety power shutoffs.
“It’s a tool we have and it’s a tool we’ll exercise,” Bill Messner, PGE’s director of wildfire mitigation and resiliency, said of the preemptive blackouts. But it’s a tool of last resort, and the aim is to use new technology to reduce the number and duration of unplanned outages, reduce fire ignitions, and isolate public safety power shut offs to smaller areas when they are necessary.
“As our climate changes, our strategy has to change,” he said. “We have to continually think about all these bad things happening. It’s nightmarish after a while.”
PGE is also using artificial intelligence and machine learning to detect fires, reducing the potential for human error, said Dan Nunez, manager of the wildfire planning and analytics. The utility has identified 10 separate high-risk fire zones in its service territory, including areas around Mount Hood, the mouth of the Columbia River, parts of Portland’s West Hills, Estacada and Oregon City.
In a modern version of a fire lookout tower, it is installing a network of 22 ultra-high-resolution cameras that capture a 360-degree view of the surrounding landscape. The cameras can detect smoke and fire ignitions, triangulate them to within 100 meters, and immediately alert local fire agencies, the Forest Service and tribes. Better response times could prevent small ignitions from becoming conflagrations. And they’re a powerful tool for the utility to monitor the condition of its equipment, with the ability in some cases to zoom in and identify an osprey nest on a utility pole 15 miles away, for instance.
PGE has also been working with Amazon Web Services to build out a wind and weather-based outage prediction model that will identify the area, specific equipment and customer impacts expected in a given weather event. It draws on many of the same data sets as the mPrest platform, and will forecast vulnerabilities down to the subsections of individual circuits. It will be used to direct crews to problem areas with the right equipment, target trimming of the 2.4 million trees within its right of ways and prioritize new equipment investments. It aims to have the platform running in 2023.
While Oregon largest utilities have the wherewithal and customer base to make these investments, the state’s 36 consumer or publicly owned utilities, many serving rural areas with high fire risk, don’t operate at the same scale. But they aren’t standing still.
Senate Bill 762 also required those utilities to have risk-based wildfire protection plans approved by their governing bodies by June 30, 2022, then submit them to the Public Utility Commission.
Tawney said those largely rural utilities, without the same economic resources or manpower, will have to find other ways of assessing their own risk and mitigating it. That may be through collaboration with the Bonneville Power Administration, the U.S. Forest Service or the Oregon Department of Forestry.
“It’s a question that worries all of us, and I keep trying to keep my ear open on the federal spending,” she said.
Dave Markham is the president of Central Electric Cooperative, a utility that serves 39,500 customers over a 5,300-square-mile swath of central Oregon, including high-risk areas around Camp Sherman, Sisters and the Ochoco Mountains.
Markham said the cooperative has been bolstering its wildfire preparedness plans since 2018. The utility submitted its preparedness plan to the public utility commission a year and a half early, and is tapping new technologies to up its own game.
The coop is using software from Texas A&M that monitors its high-risk distribution lines and can pinpoint specific issues, such as a cracked insulator that could bleed electricity onto a pole and start a fire.
It is also one of two utilities – the other is Consumers Power in Philomath – working on a research project with the U.S. Department of Energy and a Texas company developing artificial intelligence for off-the-shelf drones. Those drones can fly utility right of ways and identify damaged equipment and problematic vegetation much faster and more accurately than a line crew could conduct visual inspections.
In the end, Markham says smaller rural utilities don’t have the resources, or perhaps the need, for the big data capabilities of their larger counterparts. Many of their wildfire preparedness challenges are more prosaic and have common sense solutions – such as the ability to promptly get permits from the Forest Service to remove hazard trees.
“To me, that’s every bit as important as implementing technology,” he said. “I feel confident in our abilities even though we don’t have millions of dollars invested in this modeling.”
– Ted Sickinger; firstname.lastname@example.org; 503-221-8505; @tedsickinger