AI Demonstration Projects Show Great Potential, but Are Other Utility Companies Ready?

Utilities and system operators are discovering new uses for artificial intelligence and machine learning to address reliability threats from load growth, according to multiple utilities and analysts.

Generative AI models have "exploded into public consciousness," according to a 2024 report by the Electric Power Research Institute (EPRI). This trend has spurred massive AI investment commitments for 2025, such as the $500 billion Stargate project in the U.S. and a €206 billion fund in the EU. Utilities are beginning to realize the possibilities these new technologies bring.

"Utility executives who were skeptical of AI five years ago are now using cloud computing, drones, and AI in innovation projects," said Jeremy Renshaw, EPRI's executive director of AI and Quantum. "Rapid adoption by utilities could make things impossible today become standard operating practice within a few years."

However, concerns remain that AI/ML algorithms could bypass human decision-making, leading to reliability failures they were meant to prevent.

"But any company that hasn't yet incorporated its internal knowledge base into a generative AI model that can be queried on demand is not fully leveraging data it has paid to store for years," said Marc Spieler, senior managing director at NVIDIA. He added that humans will still be involved in decision-making for now, and AI/ML algorithms help make better decisions by providing more relevant data faster.

In practical demonstrations, utilities and software vendors are using AI/ML algorithms to improve tasks ranging from nuclear plant design to electric vehicle charging. But utilities and regulators must confront a dilemma: how to make proprietary data more accessible to new digital intelligence to improve reliability and lower customer costs, while also protecting that data.

Old technology finds new life

Power systems have already deployed AI/ML algorithms in cybersecurity applications, using advanced learning capabilities to better identify attackers.

Checkpoint Software, a security service provider for NVIDIA, the global AI chip maker, is working with standards certification body Underwriters Laboratories to develop new security standards for consumer devices, said Peter Nicoletti, Checkpoint's global chief information security officer. Smart devices "will be required to meet security standards that prevent hackers from attacking during software updates," he said.

Another proven advanced computing application for power systems is market price forecasting based on weather, load, and available generation.

Amperon has used AI/ML algorithms for weather, demand, and market price forecasting since 2018, said co-founder and CEO Sean Kelly. But Amperon's short-term modeling now "runs every hour and continuously retrains smarter and faster with less energy, integrating the strengths of each iteration in ways humans cannot match," he added.

Hitachi Energy's Nostradamus AI forecasting tool uses the latest AI/ML capabilities, "and since November, price forecast accuracy has improved 20% over manual forecasts," said Jason Durst, general manager of Hitachi Energy's asset management and enterprise software solutions.

AI/ML-assisted technology has also "become a key pillar of wildfire mitigation strategies," said Rob Brook, senior vice president and managing director for the Americas at forecasting software vendor Neara. The technology helps utilities identify "wildfire risks in their networks by proactively assessing more variables than humans can integrate," he added.

Over the past year, AI/ML algorithms have accelerated the deployment of solar construction robots, said Deise Yumi Asami, developer of AES's Maximo robot. The six months previously required to retrain Maximo have been eliminated because its AI/ML algorithms now learn the unique characteristics of each solar project autonomously before it begins.

New, more autonomous AI/ML capabilities will provide "stability, predictability, and reliability at scale," said Nate Melby, vice president and chief information officer at Dairyland Power Cooperative, a Midwest generation and transmission cooperative. Managing system complexity "is exactly where AI can shine," he added.

Utilities are increasingly leveraging new AI/ML capabilities to address growing power system challenges such as variable load and the surge of distributed energy resources.

New demands, new capabilities

Pacific Gas and Electric Company (PG&E) noted in its 2024 R&D strategy report that a power system lacking flexibility "could lead to decreased reliability and safety, increased operating costs, and capacity costs." The report added: "AI/ML and other new technologies can not only enhance our immediate response capabilities, but also inform long-term planning and policy-making."

PG&E CEO Patti Poppe said at a November innovation summit that the utility's total electricity usage will double over the next 5 to 10 years, but peak load growth can be limited to 10% through AI/ML-based grid optimization of distributed energy resources.

NVIDIA's Spieler said that access to AI/ML algorithms is now commercially viable, and their capabilities can optimize multiple large scenarios in parallel to support decisions involving millions of variables in power systems. These algorithms can also write software code, enabling utilities to leverage "their stored petabytes of system data that were previously not used to optimize more operations," he added.

Spieler said utilities can upload internal knowledge bases, including research papers, rate cases, and analyses of wildfire and safety issues, to generative AI models and query them. Query responses can explain system anomalies based on performance and maintenance history, or provide data and precedents needed to write general rate cases and other regulatory filings.

Utility demonstration projects are validating new AI/ML capabilities.

From distributed energy to nuclear plants

Multiple demonstration projects focus on how AI/ML algorithms can optimize distribution system resources.

Utilidata's Karman software platform and NVIDIA GPU chips are embedded in Aclara smart meters and will soon be applied to other distribution system hardware, said Yingchen Zhang, Utilidata's vice president of products. Karman reads high-resolution distribution system raw data 32,000 times per second and identifies individual customer electricity usage in real time, he added.

A practical demonstration found that with Karman reading and responding to real-time data, utilities can quickly stabilize voltage fluctuations caused by electric vehicle charging, according to a University of Michigan-Utillidata study.

Within one year of implementing data disaggregation expert Bidgely's software, Avista Utilities saw a 27% reduction in service calls due to high bill complaints, reported Andrew Barrington, product and services manager at Avista Corp. Bidgely's analytics identified customer behaviors causing bill spikes rather than dispatching service personnel to check meters, he added.

A Bidgely disaggregation analysis evaluated electric vehicle charging for 10,000 Ameren Missouri customers, said Caroline Cochran, its vice president of delivery, in a report at a Stanford-EPRI conference. The analysis identified 73 customers who could avoid or defer expensive infrastructure investments needed to manage EV charging load through better management, she added.

A similar Bidgely disaggregation analysis of 100,000 NV Energy EV charger owners in 2023 identified "hot spots that may warrant priority infrastructure investment," thereby limiting broader distribution system capital investments, according to a Smart Electric Power Alliance January report titled "AI Insights Brief for Transportation Electrification."

AI/ML algorithms are also playing a role in improving the cost-effectiveness and safety of nuclear plants.

PG&E is using Atomic Canyon's generative AI software, trained to Nuclear Regulatory Commission standards, at its Diablo Canyon nuclear plant, said Patrick White, research director at the Nuclear Innovation Alliance. Innovative AI/ML-based plant design, operations, and predictive maintenance are limiting costs and improving plant safety, he added.

However, utilities and vendors recognize that to more fully leverage accelerating AI/ML capabilities, utilities still need to take certain steps.

Challenges facing utilities

Effectively harnessing the benefits of AI/ML algorithms begins with recognizing their potential and acquiring and using the right hardware and software, utilities and third parties say.

Avista's successful adoption of third-party AI/ML "started with a mindset," Barrington said. Key questions are "how to enhance customer engagement, how to integrate customer data with system operations, and how to enhance system visibility and enable proactive strategies," he added.

AI/ML algorithms now extract real-time data and offer actionable recommendations, said Utilidata's Zhang. But "utilities cannot leverage these recommendations because they lack the corresponding technology and communication ecosystem," he added.

Utilities need communication technology, advanced metering and edge computing infrastructure, and data processing and storage technology, said EPRI's Renshaw. At the distribution system level, utilities should also have software that can be safely updated to accommodate new technologies adopted by customers, added Utilidata's Zhang.

Balancing safety and customer privacy protection with providing data to train AI/ML algorithms remains a significant challenge.

Protecting utility data requires "robust cybersecurity practices," said Dairyland Power's Melby. But utilities need to access and manage data "in ways that AI platforms can leverage," he added.

Recently, "utilities have begun conducting penetration tests to prove that data is as secure in our systems as in their own," said Bidgely's Cochran. They have also "formed AI committees to conduct additional thorough reviews of their data users," she added.

"Utilities are conservative about data privacy for good reason, but AI/ML power system applications have not yet posed a threat," said Utilidata's Zhang. Federated learning or foundation models are methods that both protect privacy and provide data for algorithm training, he added.

Federated learning allows utilities to protect proprietary data by building synthetic models of their specific challenge data, which can be shared in secure locations for further training, Zhang said.

But some believe federated learning may be too limited to handle the complexity of power systems. Foundation models would use orders of magnitude more data, pre-trained with anonymized data and as much power system information as possible, said EPRI's Renshaw and others.

Utilities might be able to create foundation models to enable shared learning and protect their data, said Quinn Nakayama, senior director of grid research, innovation, and development at PG&E.

"The bottom line is—collect more high-quality data, use, store, and protect it correctly, and feed it into models trained and updated for the right tasks," Renshaw concluded.