Last September, two months before the launch of ChatGPT, artificial intelligence (AI) was already helping keep the lights on for thousands of Californians during a historic heatwave.  

As the state cooked, an AI-enabled Virtual Power Plant (VPP), a coordinated network of decentralized energy resources that can operate in lieu of a physical, fossil generation power plant, was adjusting smart thermostats in homes and drawing power from residential and commercial energy storage, microgrids and electric vehicle (EV) fleets to help prevent blackouts.         

While the recent acceleration of AI development has elicited a mix of wonder and apprehension, its emerging potential to support America’s challenged power grid should inspire a healthy dose of optimism. 

The California example demonstrated this technology can not only enhance the power grid when it is under strain, but it can do so at scale. (The Virtual Power Plant in California was operated by AutoGrid, a cleantech energy solutions company owned by Schneider Electric.)       

The increased strain on the grid from extreme heat, wildfires and extended drought, combined with surging demand from EVs and other efforts to electrify everything from heating to manufacturing, suggest we’ll need new technologies and solutions to fortify the grid.  

Scaling VPPs is the major task at hand and AI has already been a game changer. The software sophistication needed to manage even a small-scale VPP is immense. In terms of data, a VPP is the equivalent of running tens of thousands of power plants in parallel, which must all operate cooperatively, responding within milliseconds of one another.  

Each distributed energy resource that contributes to a VPP — whether it’s a smart thermostat, water heater , grid-scale battery  or an EV — has its own features and limitations. Each resource’s capacity is also dependent on the expectations and usage of each consumer or business participating in the VPP.  

For instance, some participants may not want their thermostat adjusted more than a couple degrees. Others may have longer commutes and have less capacity to contribute power from their EV battery. The system must account for all those complexities.  

When we add EVs to the equation, which can both store and push power (meaning they both draw energy and can return that energy to the grid) the vehicles can help address reliability challenges, particularly during peaks in demand. Accounting for EVs and their impacts also involves a massive amount of data that needs to be interpreted and harnessed so that these resources work in concert with each other to form a system that can react in real time to the immediate needs of the grid. As VPPs scale, the complexity of this task increases exponentially.  

All of this is a challenge tailor made for AI. AI efficiently manages all of this complex data, identifying patterns, correlations and anomalies to make real-time management decisions for each energy resource to support the grid. This at-scale, rapid-learning and forecasting is far beyond human capability.  

Of course, as with all AI applications, particularly when applied to critical infrastructure, there’s a risk of bias and faulty assumptions. Caution and care are essential to building these AI models so that they don’t inadvertently contribute to power problems instead of solving them.  

As AI unleashes the growth of VPPs, I’m particularly excited about the outsized role EVs can play in demand response.  

I started my career as an engineer on a General Motors factory floor. As an executive at GM and Saturn later, I was part of the team that helped launch the first mass-produced electric vehicle, the EV1, in the late 1990s. That particular venture didn’t work out, but fortunately, the world looks very different today. Rapid EV adoption is projected to have a huge impact on overall power consumption in the next decade, with electric vehicles projected to consume 7% to 11% of all electricity generation in the United States by 2030.  

Mass EV adoption is an enormous grid challenge, but it’s also an enormous opportunity. Of the many distributed resources that contribute to a VPP, EVs are the only movable assets. We’ve barely started to wrap our minds around the implications of this mobility.  

For example, if extreme weather leads to a power outage, EVs could be deployed to a hospital or crisis shelter (school, library, community center) to provide battery support in place of diesel back-up generators.  

Nissan has already piloted this approach with its “Re-Leaf” program, and other similar projects are in the works, including a partnership between Schneider Electric and the Oakland  (California)  Public Library to pilot the use of electric buses to power air conditioning and critical services during heat waves and blackouts.  

It won’t be long before EVs are nearly as ubiquitous as cell phones. As more commercial and industrial sites convert to electric transport and more households embrace being “prosumers”—producing and managing their own energy through apps, rooftop solar, storage, bidirectional EVs and other innovations—rolling blackouts and appeals to conserve power could soon be behind us.