People first: Human insights on artificial intelligence in power plants

By Rod Walton, Power Engineering and POWERGEN content director

AUSTIN, Texas—Artificial intelligence and other digital tools promise to help remake power plant operations into something infinitely more efficient, sustainable and intuitive for the fast-forward and cleaner energy future.

But there’s one thing that can’t be artificial about it—who runs it.

Ignoring the human element in digital transformation will slow adoption of highly valuable technologies which need context to be used properly, said Steven Ford of Southern Co. Ford, who is compliance and support manager for utility and power generator Southern Co., was speaking at the Connected Plant Conference (CPC) held live in Austin’s Renaissance hotel this week.

AI, virtual reality, analytics, remote operations, predictive maintenance are not just buzzwords but true visions for plant operations of the future. Without human talent and institutional knowledge, however, these are just countless bits of information which can either save millions by fixing a thorny issue or create a massive new one.

The Connected Plant Conference, sponsored by Power Magazine and Chemical Engineering Magazine, brought together a host of expert speakers on data analytics and digital tools. They are all proponents of the ways intelligent machine operation can benefit industry, but like most of the best prophets they also acknowledge the dangers involved.

“AI is not the answer; it is an answer,” Lloyd Colegrove, an independent digital transformation consultant and retired data services director with Dow Chemical, pointed out during his CPC presentation. “It shouldn’t run a plant but you can use AI to make it run better.”

To illustrate his point, Colegrove placed small plastic balls—representing data points—on the table in front of every attendee. They then aimed at a small bucket—meant to symbolize a spreadsheet catching key data—on the stage. Attendees threw the balls which almost invariably missed the bucket.

But what if the most important ball, or bit of data, fell way outside the spreadsheet bucket and was missed by the analyst? “There is so much data, so much information flying at plant engineers, they can’t possibly do it,” Colegrove said. “You have to have a system that allows your people to be successful.”

Indeed, developing a digital twin of combined-cycle gas turbine plant, for instance, can prove highly valuable to finding anomalies in performance data and then predicting and fixing breakdowns before they happen. A bearing is a bearing is a bearing is not necessarily true anymore, as one speaker, Slavek Zaremba, chief technical officer for predictive maintenance solutions at SparkCognition, pointed out.

Making sense of the data and providing the operational context an experienced hand can give is key. How can you catch an anomaly if you don’t know enough about the plant output to know that dataset is off-base?

Marco Sanchez, vice president and head of intelligent solutions with turbine and plant manufacturer Mitsubishi Power, asked the CPC crowd to give an estimate on how many times the average power plant trips offline each year. Seven, said one attendee; 27, guessed another.

Fifty to sixty times a year, Sanchez said, altogether resulting in whistles of surprise at the event but, more importantly, in thousands of hours of lost productivity and grid response challenges around the U.S. and world. The majority of those trips are coming from balance of plant systems outside of the gen-set, and Sanchez pointed out that his company’s gas turbine is operating at 99 percent reliability.

He talked about Mitsubishi’s TOMONI suite of digital solutions, which now includes remote operations monitoring for power plants globally.

The Japanese power generation firm also has ramped up operations at its T-Point-2 testing power plant near its Takasago Works facility in its home country. T-Point-2 works to validate technologies in the gas turbine sector, including Mitsubishi’s plan to deliver a hybrid hydrogen gen-set for Utah’s Intermountain Power Project running on 30-percent H2 and eventually 100 percent in coming decades.

“It’s not a science project any longer,” Sanchez said. “It’s becoming a reality.”

This new virtual reality needs the real human touch now as much as ever. The power plant workforce is finally getting younger after aging for decades, but this change pits two challenges at odds—one is a new workforce which expects automated bells and whistles to get the job done and knows how to use them, and yet lost plant knowledge through retirement and departure of longtime veterans.

On-site plant knowledge is a must. Successful asset management of both mechanical equipment and the digital tools are part and parcel of the same thing. Terms like digital transformation sound great in a sales pitch but they can confuse and discourage employees who either want to cling to the tried and true ways or believe in the next-gen tools too much to question the results.

“All these processes (used in the past) are still good; it’s how we apply the technology to the process,” said Bernie Cook, executive vice president for Woyshner Services Co., which works with utilities on power generation and transmission projects. “It takes time to adapt to these new technologies.”

Innovation must be forward looking and fearless, Southern Co.’s Steven Ford noted, pointing out the light bulb was invented before electricity was even widely available. At the same time, companies validating the benefits of next-gen, AI-level tools should work on them in small groups and curry favor with sponsors in the executive suite. These things are exciting, but they need vision and true valuation in functionality.

Listen and learn. Communicate.

“Growth requires change,” Ford acknowledged, yet “change requires people. People get excited about new ideas when you sit down and explain them. People inherently want to learn.”

The Connected Plant Conference continued through today.

No posts to display