EPRI’s reverse pitch event is part of a larger effort to tap the power of AI to benefit the grid and society
Successful Silicon Valley entrepreneurs have a knack for pitching their business ideas. A lot of ingredients go into a good pitch. Having an impactful product or service that solves a big problem is fundamental. But successful pitches inevitably require entrepreneurs make educated guesses that the solutions their businesses have labored to develop uniquely address a challenge their audience wants to solve.
That traditional paradigm for matching business solutions with business challenges was completely upended at EPRI’s recent Reverse Pitch event for AI and electric power companies.
“When you have solutions that are chasing problems, you often get the situation of trying to force a square peg in a round hole,” said Jeremy Renshaw, an EPRI senior program manager who leads the AI.EPRI Initiative, which seeks to promote collaboration among utilities, leading artificial intelligence (AI) companies, and academic researchers. “We wanted to bring the problem to the solution. So we had utilities bring forward data challenges and present them to AI companies and researchers instead. Then the AI community responded with solutions they have or could develop to address these—and potentially other—challenges.”
The event brought together utilities like Duke Energy, RTE, TVA, PG&E, PPL, Ameren, and Southern Company, universities such as Stanford and MIT, national laboratories such as Lawrence Berkeley National Laboratory and the National Renewable Energy Laboratory (NREL), as well as AI companies like Intel, Google, and C3 AI. Over 100 organizations participated in the event, where AI experts responded to utility reverse pitches about decarbonization, asset management, digital twins, cybersecurity, grid resiliency, and outage optimization.
For example, Kevin Thompson, a manager for Asset Information and Intelligence for Duke Energy, says he was eager to participate in the event to learn more about how AI and machine learning can help improve the maintenance and replacement of grid assets. “We see huge potential for AI to move us from time-based to condition-based maintenance and replacement so we can reduce failures,” said Thompson.
Facilitating Collaboration to Accelerate AI Adoption
The reverse pitch event was one of many gatherings EPRI organized to increase collaboration between the two industries and was co-hosted by Stanford University’s Bits & Watts Initiative—the university’s effort to accelerate technology innovation to benefit the 21st century grid.
Earlier in 2021, EPRI, Stanford, and NREL hosted the event This is AI: Introductory Training Course and Expert Panel on AI for Electric Power Experts, which featured presentations by Google, NREL, and Stanford. Another was the virtual roundtable Convening AI and Electric Power, which included participants from AI companies like Microsoft and RWI Synthetics, utilities such as Ameren and CPS Energy, and NREL. In September, EPRI will host its first AI and Electric Power Summit, which will focus on how AI can accelerate innovation for a clean energy future.
In addition to these events, EPRI researchers are identifying applications for AI in the power industry and educating utilities about a range of AI topics, including technical topics like natural language processing and image processing as well as issues around use cases, data sharing, and data governance.
Such collaboration and education are critical for empowering utilities to tap AI’s potential value. “The utility sector is very risk-averse. A lot of utilities have applied AI to some areas, but they have not yet realized the full value,” said Liang Min, the managing director of Stanford’s Bits & Watts Initiative. “We are trying to form a consortium of academia, national labs, utilities, and AI companies to de-risk the adoption of AI technology and accelerate research, development, and deployment of AI.”
Grand Challenges to Focus Collaboration
At the AI and Electric Power Summit, EPRI and its collaborators in the AI and utility industries will unveil and discuss a series of grand challenges—specific areas where AI can make a meaningful impact on utility operations and objectives.
Grid-Interactive Smart Communities: As more homes and commercial buildings connect to the power grid, AI can be used to improve communications, energy efficiency, load shifting, and emissions reductions among a vast array of distributed energy resources to support decarbonization.
Energy System Resiliency: AI could help predict weather, electricity demand, and plant and grid conditions. This would allow the power system to continuously optimize in ways that minimize the unplanned outages of assets. AI could also intelligently control energy flows to minimize or eliminate future extreme weather impacts and reduce unplanned outage durations.
Environmental Impacts: AI can help maximize the grid’s use of renewable energy, reduce wildfire risk, improve vegetation management, minimize wetland impacts, and help protect endangered species.
Intelligent and Autonomous Power Plants: Resource flexibility is becoming increasingly important as cleaner, distributed energy resources are added to the grid. AI can automate power plant startups and improve maintenance so that central station resources connected to the grid can work together efficiently.
AI-Enhanced Cybersecurity: AI can be used to monitor the power system’s system’s information technology (IT) and operational technology (OT) components and systems to identify suspicious activity and respond to potential cyberattacks faster than humans can and help to support human cyber experts.
Collaboration to Unlock the Potential of AI
While the grand challenges are distinct and require unique solutions to address properly, they all share something important. They are critical to the success of the future power system and represent significant enough challenges that no single utility or organization can solve alone. Collaboration across industries and companies is a necessity.
One of the presenters at the reverse pitch event was Steve Orrin, who leads Intel’s technology engagement with the government. Orrin discussed AI’s potential to identify cyber attacks on the grid more quickly and also serve as a tool to improve grid resilience. “Good AI can understand normal and abnormal grid behaviors,” he said. “It can also prevent outages because the anomalies it detects could be related to a fault that can be addressed with maintenance.”
Orrin emphasized that for utilities to derive value from AI, they need large, high-quality data sets along with rules that protect that data. “The hard part is not developing AI models; it’s getting the right data to drive the models,” he said. “The number one barrier to improving cybersecurity is getting access to data from IT and OT.”
According to Orrin, before utilities can apply and reap the benefits of AI, large datasets need to be collected, curated, protected, labeled, and shared. It’s those large data sets—which may be impossible for any single utility or organization to collect—that allow AI models to be built and to improve continuously. EPRI has assembled ten datasets, called the EPRI10, spanning topics from maintenance to operations to power quality and even satellite imagery that utilities can use to build AI models.
While utilities are constrained by strict regulations related to customer data, Orrin says that there are precedents that utilities can follow. “This requires governance. You don’t just publish IT and OT data. You need to anonymize the data and compile it to create models,” he said. “We’ve seen how to do this in financial services and in government.”
Duke Energy’s Kevin Thompson says that he is particularly interested in EPRI’s efforts to advance power industry-wide data sharing and collect high-value datasets that can accelerate the industry’s use of AI. “We want all utilities to share their data so that they can leverage AI and machine learning,” said Thompson. “By ourselves, we don’t have enough data. We need other utilities to join us in this effort and to structure their information in the same way. If I call something an apple and you call it an orange, we can’t work together.”
Duke Energy already uses AI to identify transformers at risk of failure and in need of additional testing and maintenance. The utility is examining the use of AI to inform decisions about replacing a wider range of aging grid equipment. “Many utilities are in the same place as us. A lot of grid infrastructure was built over 50 years ago and can’t be replaced all at once,” said Thompson. “AI and machine learning can tell us the risk of equipment failure, which can be more effective than the historical approach of replacing equipment based simply on how long it has been operating.”
Key EPRI Technical Experts:
For more information, contact email@example.com.