A Coal Refuse Power Plant Is Pioneering an AI-Driven Overhaul

An iconic coal refuse power plant in West Virginia that burns gob—basically dirt—and pond fines as part of a complex environmental solution to address coal waste is stepping up efforts to improve its efficiency, reliability, and cost reductions by embracing cutting-edge artificial intelligence (AI).

American Bituminous Power’s (AMBIT’s) 80-MW Grant Town Power Project—West Virginia’s only remaining coal refuse–fired facility—is partnering with AI software and advanced robotics firm Gecko Robotics to leverage data-collecting robots, data integration, analytic tools, and software modules. Under a three-year partnership revealed to POWER, the companies will deploy Cantilever, Gecko’s newly launched AI-powered platform, to drastically reduce forced outages at the unique plant.

The partnership is notable given its commercial scope and applicability, even in an industry that appears to be rapidly adopting AI tools. Only three years ago, the power sector struggled to uniformly agree upon a definition for AI, the broad concept that encapsulates how machines can imitate human intelligence. Today, AI facilitates and enables new approaches, from predictive maintenance, streamlined work, grid management, risk evaluation and detection, load forecasting, failure detection, efficiency, and emissions reductions.

Industry progress, however, has been irregular, hampered in part by investment and outcome uncertainty and a cautious approach to power plant modifications. Gecko’s partnership with AMBIT is slated to “show the way as it relates to how to actually make sustainability, availability, and resilience a science, as opposed to some sort of art,” Gecko Robotics co-founder and CEO Jake Loosararian told POWER.

“With the introduction of Cantilever, we are bringing software and AI into the equation to supercharge this effort and give plant managers incredible clarity. There is no more guesswork—we can predict exactly what will fail, automate repair plans using AI to maximize budgets and increase the useful life of customer infrastructure,” he said.

An Intelligent Integration for Crucial Plant Insight

Cantilever, which Gecko formally launched publicly only three months ago despite deploying it for about a dozen customers, is a turnkey asset management solution. It is powered by more than 10 years of inspection data captured by Gecko’s robotic fleet—its core business—from pipelines, boilers, tanks, and ship hulls, Loosararian explained. The data is then integrated with other data sources, such as maintenance records, operations history, and other business data.

“So, it’s a combination of all the data collected from the robotics and sensors, plus an addition of software modules and data processing.” Bringing together the data layers allows human experts to contextualize the data, provide a comprehensive view of asset health, and turn it into action, he said.

“We’ve used Cantilever with customers in the manufacturing, energy, and defense sector,” Loosararian said. “The decisions that are being made with it are actually taking information and data that we either gather,” either from the client or historical information gathered from Gecko’s assets, “to actually hike into Cantilever’s modules,” he said.“The AI-powered solutions, the modules and software, allows [clients] to see what things are going to fail and when they are going to fail because of the historical data.” The technology can also ensure that critical infrastructure facilities are deployed as planned, and with asset longevity in mind, he noted.

In yet another application, customers are using the information to pinpoint exactly what materials and solutions repairs will require. “For one of our naval customers, [repairs] were three to four months faster than previous methods using a traditional manual inspection and manual process,” Loosararian said. “So it just enables software experts to be way faster and more accurate at their job while ensuring you’re not wasting billions or hundreds of millions of dollars in the wrong areas.”

A Focused Approach to Reliability

The technology, Loosararian noted, addresses a critical gap in the power industry. “Resilience and reliability are paramount, but you cannot have resilience and reliability if you are uncertain that the infrastructure we rely on will actually perform when we need to use it as a service. And so, one of the most important things our technology is doing is addressing that aging infrastructure,” he said.

Over the course of its decade-long operations, Gecko has gathered or modeled information on a wide variety of components spanning “thousands of different types of boilers, thousands of tanks, pressure vessels, and piping,” Loosararian suggested. “To be able to have enough information to build a software that is able to leverage data that was never before possible, and then leverage that to ensure resilience, reliability, and sustainability. Those things are data-driven as opposed to reactionary,” he said. “We intend to take the grid from a reactive stance to a proactive stance.”

Grant Town, specifically, has made for an ideal first partner because the facility recognized the return on investment (ROI). That underscores a necessary “shift in mindset” that the industry should pursue while exploring digitalization solutions, said Loosararian. “Instead of reacting, which is, you typically spend way more when you react,” more actionable focus should comprise “predicting and preventing,” he said.

Grant Town’s Maintenance Boost

At Grant Town, as with many smaller-scale power plants that combust alternative fuels, the stakes are particularly high. The plant, which POWER recognized as a Top Plant in 2021, is sited on 340 acres at the site of the former Federal #1 mine in Grant Town, Marion County, given its environmental stewardship. Since 1993, the facility has remediated 530,000 tons of coal refuse annually and reclaimed roughly 30 acres of land per year. In addition, the plant produces 660,000 MWh under an electric energy purchase agreement, which ends in 2036, with First Energy’s Monongahela Power Co.

The plant relies on coal waste trucked to the facility from reclamation sites using two Foster Wheeler/Ahlstrom Pyropower circulating fluidized bed (CFB) boilers to create the steam to drive a turbine generator. Depending on availability, AMBIT typically tries to blend a 50-50 mix of gob (which has a Btu content as low as 3,000 Btu/pound) and pond fines (ranging from 7,000 Btu to 9,500 Btu) to achieve a given fuel quality each day.

However, Grant Town’s CFB boilers are “subject to a high level of erosion on external furnace surfaces, the water wall tubes, the backpass tubes, and all that,” Steve Friend, AMBIT plant manager, told POWER. “From an environmental standpoint, the challenges that we face are exactly the same as they were before. We burn waste coal, a highly variable product that can vary in sulfur content and heat content of the fuel,” he noted.

“Those are everyday challenges that we face to keep the plant running at a high level of reliability. But from a mechanical standpoint, probably one of the biggest challenges faced by the boiler is the erosion of the water wall surfaces within the boiler itself,” he said.

In 2020, after a spate of forced outages that diminished its annual power production and decreased revenues, Grant Town took up Gecko’s offer to perform robotic inspections on their two boilers. After subsequent manual inspections provided “limited data and missed damaged areas of the boilers that led to failure, in 2023, the power plant team “decided they needed to return to technology to ensure reliability,” Gecko said.

The approach deployed Cantilever, starting with a full-coverage robotic-inspections assisted analysis that “transformed data into interactive digital twins and reports in the software platform,” the firm noted. The models, which “revealed where urgent repairs were needed,” ultimately allowed Grant Town to repair and replace damaged boiler tubes. The entire process was completed during a five-day planned outage, it underscored.

Using Gecko’s robotic data collection, Grant Town is able to “get thickness measurements on 100% of the surface within the combustor of the boiler,” Friend explained to POWER. The software works by assessing tube thickness in correlation with wastage rates on the tubes’ external surface, which then allows for an accurate prediction of when they could fail, he noted.

“If the plant were to take manual ultrasonic thickness readings on the tubes at various locations around the plant in our boiler, we just don’t get that density of fitness measurements through it,” he said. “So we get 100% coverage of the health, the water wall tubes, within each boiler by using their robot collection.”

But beyond that, “using the software modeling tool that allows us to predict the remaining life of virtually every tube within the furnace,” Friend added. “That allows us to do a much better job of planning for future outage work, future replacement work, and to more accurately address problem spots while we’re in an outage to make sure that we’ve got the reliability level at the highest it can possibly be.”

Reliability and Costs Benefits

Friend underscored the value of the three-year partnership with Gecko. “From my perspective, what the three-year agreement brings to us and brings to the table is it will have three years worth of data and three years of opportunity to plan repairs and execute those repairs,” he said. “So, hopefully, at the end of the three-year partnership, we’ve got the furnaces back to a condition—won’t be new—but they’ll be in a condition to run at a high level of reliability and hopefully significantly reduce our forced outage rate due to combustion water wall leaks.”

The cost benefits may also be significant, Friend suggested. “If I were to take a manual ultrasonic thickness collection effort and cover every inch of the two [boilers], first of all, it’s an unrealistic approach because the time it would take to do that just doesn’t fit within our realm of reality for Grant Town during a planned outage period,” he said. “And secondly, the amount of manpower would be exorbitant from a cost perspective.”

Gecko serves Grant Town with a “minimal crew.” During two outages—one on each boiler—this fall, “they had a series of four robots set up, and I think five or six people there to run these robots,” Friend said. “We get data collection and data reports back virtually within a 24-hour period, if not less, so that my welding contractors can follow up behind them in almost no time, to where we need to point them to repair areas or replacement areas.”

What it boils down to, Friend said, is “Having the data collected digitally, and the data density that goes along with it brings a lot of value to us in that we get a very clear picture of what the wastage rate is on every inch of every tube in the furnaces of the boiler. Plus, the ability to be able to predict where potential failure areas are going to come from,” he said. “In the future, it gives us a huge advantage in an outage planning project.”

Friend noted that Grant Town also utilizes another AI system added to its furnace controls and its combustion controls on the boiler that uses plant data and applies AI to provide advanced process control within the combustion process. “ A simple way to explain it is that it’s like a fleet of smart cruise control for your car. We’ve developed a set of very detailed and extensive models on how the boilers operate that allows the control system to make predictive adjustments based on those models and actually tighten up the combustion process.” The benefit has been an “improved lifecycle and utilization with an improved sulfur emissions control.”

“So, it kind of goes full circle,” he said. “We’re addressing the reliability side from the maintenance perspective using the enhanced [ultrasonic testing] tools that Gecko provides. And then from an efficiency standpoint, we’re applying advanced process control to try to maximize plant efficiency, thus minimizing our operating cost.”

—Sonal Patel is a POWER senior associate editor (@sonalcpatel, @POWERmagazine).