In the renewable energy era, the ‘smart grid’ is going to have to be unbelievably smart

January 16, 2020

Associate Professor Wencong Su explains why retrofitting the electric grid to accommodate electric vehicles and wind and solar power is such a complex challenge.

Associate Professor of Electrical and Computer Engineering Wencong Su
Associate Professor of Electrical and Computer Engineering Wencong Su

Associate Professor of Electrical and Computer Engineering Wencong Su has spent much of his young career ahead of the curve. For example, his dissertation research on how large-scale use of electric vehicles could impact the reliability of the electric grid was so prescient in early 2010, he sometimes had trouble selling the significance of the work. “At that time, the penetration of electric vehicles [EVs] was so small. EV chargers were still a novelty,” Su says. “So even people in the industry, nobody saw this as a big problem.”

How times have changed. Today, electric vehicles are seen as the auto industry’s fast approaching future — one which, just as Su forecast, is thought to have huge repercussions for the electric grid. And Su says that’s just one potentially grid-disrupting technology. Once renewable power like solar and wind start to make up a sizeable share of the energy we consume, the system will get a lot more dynamic, and therefore, unpredictable. Basically, how much — and interestingly, when — we consumers use energy could look a lot different in the relatively near future.

The grid is heading for a major paradigm shift

To understand why, Su says it’s helpful to first understand how the grid works today — especially how predictable it is. In technical terms, Su says our energy system is what’s called a “load-following” system, which essentially means that our electric utilities produce energy at mostly large-scale facilities in response to what the “load” or demand is. The trick to making it work is knowing what that load is going to be on any given day and time of day. And Su says with enough long-term usage information — and additional data like weather forecasts — that’s actually pretty easy to predict. Even for a major metropolitan area, the demand forecasting margin of error might only be 2 or 3 percent.

That predictability basically gets thrown out the window, though, when you start adding lots of renewable energy and electric vehicles to the mix. For starters, renewable energy is far more intermittent than today’s fossil fuel-based power: If the wind slows down unexpectedly for 15 minutes, or some clouds temporary shade out a solar array, that means less energy suddenly streaming into the system. Moreover, to produce enough energy, we’ll need lots of renewable power plants, each of which will be far smaller and more dispersed than today’s power plants. For example, in this system, even a homeowner’s solar array feeding energy to the grid is functionally a small power plant. 

This new setup unfortunately introduces thousands of points where the power-producing part of the grid can temporarily go down. It’s a similar portrait of unpredictability, Su says, when you start introducing EVs en masse: “When it comes to demand, no one knows what putting hundreds of thousands of electric vehicles out there will do to the grid.”

The grid of the future

Given these challenges, the big question for researchers like Su is whether the “always on” predictability of the grid can be preserved in an era of renewable energy and widespread use of EVs. Su thinks the answer is probably yes. But it’s a tricky puzzle to solve — in part, he says, because the solution to the reliability problem will likely be a patchwork of many smaller solutions. And the real magic, will be getting all of them to work together seamlessly.

One likely piece of that puzzle is a subject Su’s been studying for years: EVs. Sure, they can be a disruptor to the grid when hundreds of thousands of cars are “refueling,” drawing energy from the grid, one at a time, in an unpredictable fashion. But those same EVs, when not in use, can also be thought of as a massive, dispersed battery storage system. Feeding energy back to the grid, EVs could actually help stabilize it if, say, a dip in wind speed caused a temporary slowdown in production. Fleets of EV school busses sitting idle all summer, for example, could basically become a small power plant.

Similarly, he forecasts the widespread, perhaps even mandatory, adoption of a practice known as “demand response.” During times of peak load — like a super hot day when everyone is running air conditioning — a utility could turn up your thermostat setting a few degrees to reduce strain on the system, thus preventing brownouts. Su says such connected systems require the consumer giving up a little bit of control, but price incentives could ease that pain. In fact, demand response is already being used across Europe and in some places in the United States. Su himself participated in a pilot of such a system as a graduate student, and its effects went unnoticed. 

Further, another one of Su’s solutions would give consumers a kind of agency they’ve never really had when it comes to electricity. It’s called dynamic retail electricity pricing, and the basic idea is that there wouldn’t be one flat kilowatt-per-hour rate for power. Instead, electricity would potentially cost less when the supply was high — say, when a sunny, windy day was flooding the grid with renewable energy. But it might cost more when the supply was short. The idea is that price could be used as a mechanism to encourage conservation at critical times; plus, it would give people more choices about when and how they choose to consume — or even produce — energy. “We can imagine a driver pulling into an EV charging station and getting one price for immediate recharging; or a better rate if the driver waits five minutes,” Su explains. When it comes to balancing supply and demand, Su says minutes could make all the difference.

As you may have guessed, the secret sauce that makes all this work is artificial intelligence. Su says deciding, for example, when to inject a pricing incentive into the system to respond to a temporary dip in supply, and to do that thousands of times per day, in real time is way too complex a challenge for human brains. Powering such complicated energy management systems could even be a tough one for machines. Su says we may yet have to see new advances in computing, like parallel computing, to handle computational problems that are as complex and dynamic as those presented by the smart grid. 

One final forecast from Su: He says no matter how the smart grid evolves, we’re likely to have much more awareness of our relationship to energy than we ever have. It may still be as reliable as ever, but the days of taking our power for granted could be numbered.


Su’s research has been supported, in part, by the University of Michigan, National Science Foundation, Department of Energy, Department of Defense and local industries.