AI Behind The Light Switch: The Future of Power Utility
The invention of the commercially viable light bulb is one of the iconic stories of 19th-century technology.
Thomas Edison developed a personal and transformative vision, ran countless experiments, and emerged with not just a working prototype but also a lasting business empire that would serve as the model for the new utility industry.
Today, however, utility companies born out of his 19th-century technologies face modern challenges that Edison, the paragon of human intelligence during his generation, could never have foreseen.
The Modern Utility
The wealth and security of information-age nations — not to mention the well-being, comfort, and productivity of consumers — now depends on the smooth functioning of the power utility grid, including protection from hackers and malware.
A retiring workforce, steeped in engineering expertise related to physical machines, is struggling to transmit its knowledge of turbine management to a new digitally driven generation.
And after a century-plus of burning fossil fuels to satisfy consumer and industrial electricity needs, many utilities are bringing new clean energy sources online, thereby facing complex integrations with renewables and battery storage.
Added together, these transitions are more than a leading team of 20th-century utility experts can manage.
Why Utilities are Investing in 21st-Century Solutions
Challenges created in a digital age call for digital solutions.
Artificial intelligence (AI) and machine learning (ML) technologies hold the problem-solving promise of a million Edisons.
“The promise of AI is that it will objectively look at a situation or problem, freed from the biases of human thinking,” said Shane Mickey, VP of Strategic Planning at Mitsubishi Hitachi Power Systems (MHPS) Americas. “Once deployed and trained, AI will expand the reach of the dwindling number of true human experts in our industry. These interactions will then yield more results and correlations which we might not have thought of and either make decisions or help us make decisions much faster and better.”
With 41% of the total installed capacity, MHPS is the world’s top manufacturer of advanced class gas turbines and a leader in power utility plant reliability and efficiency through the use of big data.
Leading the Charge
The company’s digital solutions program is called MHPS-TOMONI (TOMONI means “together with” in Japanese in the spirit of collaboration). It was created in an effort aimed at envisioning and putting into action the digital power plant of the future.
MHPS-TOMONI is on pace with, if not a step or two ahead of, industry trends toward incorporating AI solutions to organize data, learn from it, and develop automated protocols that detect and address plant problems as they develop in real time.
“In the past, a typical interaction would consist of a component within a plant alarming when it sensed conditions were getting dangerous for that piece of equipment or people,” Mickey said. “It would then depend on human interaction to diagnose the alarm condition and take action.”
With AI, these diagnoses and decisions are made instantaneously, meaning equipment stays online and consumers will see fewer, shorter periods without power. The application can also learn from its own experience to ensure it always reacts optimally to changing conditions.
“There is no doubt that AI is going to be a game changer, in the same way that the internet was twenty years ago,” Mickey said, “but it feels like we’re in the early days of AI.”
As with internet-related investments in the late 1990s, smart AI moves promise to pay significant dividends, but it can be hard for those new to the technology to separate the bona fide from the bogus.
These days, forward-thinking utility executives find themselves searching for answers, hoping to find the right solution to deploy AI to solve emerging, industry- wide challenges.
The Problems AI Can Solve
When asked to identify emerging challenges facing power plants in the near future, Mickey mentions two that will be familiar to any utility executive — both of which can be addressed by adopting AI solutions.
“The power industry faces two macro-trends,” Mickey said. “The first is an aging workforce. How do we transfer knowledge of our retiring workforce to fewer people in our industry so that future power plants are more reliable than today’s power plants? That’s no small challenge.”
The second industry-wide trend, according to Mickey, is the adoption of renewable energy sources and technologies.
“Battery storage has the potential of making renewables relevant when the sun doesn’t shine and when the wind doesn’t blow,” Mickey said. “This future mix of distributed power assets, along with central power stations, will make it much more challenging to maintain a stable, reliable electrical grid.”
Both issues are on the radar of AI developers in the utilities space. In fact, AI applications can be seen as a potential replacement strategy for retiring subject-matter experts — especially since there are few candidates willing to gain deep expertise in the field.
Rethinking the Grid
With the power of AI, algorithm builders are training computers to capture historical knowledge, predict the future, and correlate future states with historical data.
Soon, computers will be able to recognize and remember situations that only a subject-matter expert or industry veteran would be able to correctly diagnose today. MHPS has already seen positive results from an AI system used in Taiwan beginning in 2016 to automate combustion tuning in boilers.
“The initial verification testing demonstrated results that are comparable to those that could be achieved by a highly experienced engineer,” Mickey noted.
As difficult as it is for utilities to lose retiring subject-matter experts, the fact is that even a long-serving expert with encyclopedic knowledge of the 20th-century power plant would be at a loss when confronted with the new challenges of a greener era.
The adoption of renewables in the electrical grid is forcing utilities to rethink many elements of their systems. Renewables depend on weather and seasons, which make them less predictable than traditional energy sources, and necessitate better predictive abilities. In this process, power plants will come to rely on AI applications as a “green Edison,” lighting the way to a more sustainable future.
Mickey sees integration with renewables as one cornerstone of AI design, along with other forms of operating flexibility.
“A plant that is AI-enabled with MHPS-TOMONI will allow central power stations to take in external and environmental conditions, such as weather or distributed resources coming on- or offline, and ramp up or down to meet demand changes,” he said.
AI and ML will transform the utility industry in a number of important areas.
Namely: cybersecurity (recognizing hackers and other threats using subtle correlations of warning signs), applications on the Internet of Things (e.g., understandable usage data to help end users determine which specific fixtures cost the most), and energy trading and risk management (e.g., better forecasting of load and risk factors to help utility pricing).
Perhaps the most exciting thing about AI for early adopters is the sense that there may be no corner of the utilities industry that AI can’t transform.
“One of the most interesting things we’re going to get out of ML is learning what we don’t know already,” said Jim Taylor, CTO of Tucson Electric Power. “These systems are going to identify issues and opportunities that we’ve never even thought about looking for. That’s the beauty of a machine being able to do it — It’s going to find things that we hadn’t even thought of and lead us to conclusions that we wouldn’t have reached on our own.”
The Early Results of AI Investment
The first question for many power companies setting out to invest in AI is whether outside help is needed at all. Large companies that attempt to build AI solutions, as opposed to buying them, will face challenges like needing to think a decade ahead of present needs in order to keep useful applications in the pipeline.
Just equipping staff to begin to build AI can be as complex a decision as buying an end-to-end solution. In short, building something as complex as AI will be an uphill battle for companies designed to compete in other industries.
MHPS-TOMONI, the partnership-driven digital power-plant project at MHPS, is a more sustainable approach built on trial and error with partner-supplied AI applications.
“We see this technology enabling the power plant of the future in general terms, but we’re going to have to experiment, in partnership with companies like SparkCognition, and our customers, to figure out the best and most cost-effective solutions,” Mickey said.
Every power plant will come to the technology with a different amount of preparation required to implement it. Some have been collecting and organizing performance data for years with rules and procedures in place.
Others may be late to the game.
A good AI solution will be nimble enough to adapt its technology at any stage of the digital transformation process.
AI is a new factor in the utility industry. For some, it may be some years before smart investors see a substantial return on their investments. Returns in the form of fewer shutdowns and cybersecurity incidents, better consumer offerings, and smoother integration with the 21st-century grid.
But Mickey is confident that MHPS will be pleased with their AI engagement sooner rather than later.
“[SparkCognition is] providing a cost-effective solution that should pay back within a couple of years in our models,” he said. “Yes, there is more variability on our models than what we’re used to in an investment, but the risk versus reward is also low.”
Mickey is convinced of one thing: investing in AI is the right call in 2018. This is true for consumers as well as utilities.
“Go back to 1995 and decide if you should invest in Yahoo or Google,” he says. “Sure, it’s an easy answer now with hindsight being 20/20, but it wasn’t so clear back then. Investing in the right technology provider was a tough call then as it is now. However, investing in the technology itself, such as the internet, was an easy call then as AI is now.”