Could EVs overload the power grid?
Image credit: Dreamstime
As nations prepare for an onslaught of electric cars on the road, we delve into the impact such demand could have on the national grid.
An £11.4m programme of funding by the Department for Business, Energy and Industrial Strategy is getting off the ground to make it easier for an electric vehicle (EV) to return electricity to the network during times of peak loading. Since June, all new private chargepoints must be smart enough to adjust their rate of charge in response to external signals, in order to ease demand on the network at times like the evening peak.
There are clear reasons as to why the British government, among others, is keen to have EVs react to the needs of the grid. Without that ability, the risk is that if too many drivers in the same area demand their cars be ready to drive in the next hour it will lead to sudden blackouts.
There are distinct differences of opinion on how disruptive EV charging will be to the grid as uptake increases. Much depends on which geographic scale you use. A 2018 report by McKinsey, which cited data from Germany, is that as a percentage of overall electricity consumption, the expected growth in EV usage will add just 1 per cent to the total generation needed. Even with massive penetration of the kind forecast by National Grid ESO (Electricity System Operator) for the UK over the next couple of decades in its most optimistic net-zero scenarios, the relative efficiency of electric traction means road users would account for a little over 10 per cent of total grid consumption by 2050.
The issue for those managing different parts of the grid is that the increase will possibly or even probably be very far from evenly distributed. The extra load at peak time could make a dramatic difference to the ability of individual feeder circuits to cope.
Marc Spieler, head of global strategy for energy at chipmaker Nvidia, argues the difficulties could emerge quite quickly in electrical distribution grids around the world if they do not manage the situation carefully. “As we get to 4 to 5 per cent penetration of EVs, I believe we’re going to start seeing some major issues with the grid when it comes to resiliency and reliability, let alone 30 per cent EVs or 50 per cent EVs, which most countries are striving for.”
The McKinsey study estimated that, using its German scenario, peak load on a typical residential feeder circuit of 150 homes with 25 per cent EV penetration would increase typical peak load by 30 per cent. Fast chargers make the situation worse because of their spikier load demands. Other studies have estimated the peak load could double.
The obvious answer is a lot more network reinforcement, but that could be expensive, especially if a large number of home installations are converted to three-phase connections to reduce the risk of high demand on one phase causing intermittent problems. Reinforcement is already part of Ofgem’s plans for the UK distribution network to handle the influx of both higher demand from heating and intermittent supply from rooftop solar panels.
A study published in 2020 by Christian Calvillo and Karen Turner from the University of Strathclyde’s Centre for Energy Policy concluded that the combination of dumb charging and a focus on reinforcement for peak capacity would increase the cost of the upgrades from now until 2050 from £40bn to £70bn.
Demand peaks at normal times are only part of the problem. Project Rev, funded by the National Grid ESO, has identified six ways in which mass adoption of EVs and home charging could trip up the grid under periods of high stress if these issues are not mitigated before the next decade, when vehicle numbers are expected to soar.
One of the issues is the clear difference in behaviour between cookers and other traditional high-energy appliances and those that use electronically controlled power converters such as EV chargers when the grid is under pressure. Ovens and kettles may have trouble delivering useful heat, but they tend to respond to voltage reductions by reducing their own power demand. On the other hand EV chargers, designed to deliver a carefully controlled current to lithium-ion batteries, are highly inelastic loads.
In a webinar earlier this year to describe the initial findings of Project Rev, Martin Bradley, consultant to electronics company Sygensys, which produced the initial Project Rev report, and former National Grid ESO dispatch assurance manager, described the core problem. “Traditionally, domestic loads have been quite benign in terms of the grid,” he said. “As the voltage goes up, their load goes up. And as the voltage falls, their demand falls. EV chargers, by contrast, will be largely constant-power loads.”
Chargers would not change their power demand even if the voltage changed dramatically, Bradley added: “So they offer no relief to the grid. It means that the models that are used to study that instability will need to evolve.”
‘As we get to 4-5 per cent penetration of EVs, I believe we’re going to start seeing some major issues with grid resiliency and reliability, let alone 30 per cent EVs or 50 per cent EVs, which most countries are striving for.’
The issue with EV chargers’ inflexibility in terms of how they connect to the grid is much the same as with distributed renewables. With high levels of rooftop solar deployment, sunny days can lead to voltage increases in local areas that increasingly will need to be compensated actively instead of relying on the natural inertia from traditional heavy electrical gear. Operators have begun to seek synthetic inertia from renewables. One example is the Power Potential R&D project run by distribution operator UK Power Networks and National Grid ESO. The aim is to let renewables operators with more than 1MW of generating capacity within a region coordinate with the operator to act as a virtual power plant that does not just produce energy but provides the necessary combination of active and reactive power to help correct problems with frequency and voltage arising elsewhere in the network.
Virtual power plants are similarly being assembled from collections of EV chargers by companies such as EV.energy. The start-up is involved with a number of trials in the UK as well as in the US, running systems that actively communicate with chargers to control when and how they activate. In the current implementation, the company is focused on smart charging or V1G, in which the software takes its cue from automated contracts made with distribution systems operators (DSOs) and others with responsibility for grid stability to change the time at which EV charging kicks in.
“We shift load away from the evening peak, freeing up capacity on the grid,” says William Goldsmith, head of grid and data services at EV.energy.
The demand moves into periods not only where the energy is more readily available but often where it can be supplied by a higher ratio of renewables. One of the data sources used by EV.energy’s software is National Grid ESO’s Carbon Intensity API, which forecasts up to 48 hours ahead how much carbon dioxide the UK grid will emit.
The carbon intensity calculation is at the lowest priority for the software. “The first level is always the driver’s requirements. They put in when they need their car charged by, and the software looks at the vehicle’s state of charge. If you have an off-peak tariff, we make sure we are decreasing your bill. The second level of priority is interacting with the grid and the energy market,” says Goldsmith. That interaction allows distribution network operators to call for demand reduction if their systems detect problems that need a rapid response.
As it stands today, the price signals that consumers can use are limited. “More than 70 per cent of users are on flat-rate electricity. If there are no events on the grid, then the priority comes from renewables,” Goldsmith explains.
Though it is theoretically possible for individual consumers to interact with the grid in this way, EV.energy has taken the approach that flexible charging needs to be easier to use if it is to be a viable method for load spreading. That, in turn, makes it easier to aggregate larger numbers of chargers into an aggregated virtual power plant that can trade as a whole, especially in the day-ahead markets where the ability to estimate supply and demand reliably is essential.
For a virtual power plant to properly earn its name, it needs to be able to export energy to the grid. That comes with vehicle-to-grid or V2G operation, the focus of numerous trials around the world and the focus of the new BEIS programme. EV.energy will take part in a new trial organised by the Energy Systems Catapult, which involves two of the few vehicle makers that currently provide support for V2G: Nissan and VW.
For consumers on time-variable tariffs, the direct financial benefits of V2G are likely to be higher as DSOs can be expected to want to call on the batteries at times of high network stress. There is, however, a reason why the availability of V2G-compatible chargers and cars remains low in the UK specifically, and that lies in the current rules for trading electricity. Ofgem has yet to make the changes to pricing and settlement, including an increase in frequency to half-hourly, that would support this kind of highly granular flexibility. “It’s also cycling the battery more. The vehicle manufacturers are very happy with smart charging as it is today. When it comes to V2G, they are very keen to avoid degrading the battery unnecessarily,” Goldsmith says.
Trials will show how much V2G can affect the battery, though the outcome is not certain to be negative. Models developed at the University of Warwick showed a possibility that benign cycling could slow down the rate of degradation, though these cycles may prove to be incompatible with real-world use where drivers need to know their vehicle will be ready at certain times.
Another study published in 2021 by the University of Rochester in New York, argued that, in aggregate, battery degradation has a surprisingly low impact on overall costs and profits from flexibility, though whether this translates to individual attitudes is an open question. It is likely that charging profiles will be vehicle-specific in order to minimise battery fade. That, in turn, may complicate the schedules and forecasts used by virtual power plant operators and DSOs.
A second complication is that the charge might not be returned to the grid directly but to home systems. This vehicle-to-home (V2H) structure may prove to be more popular in regions where outages are common and so consumers expect to create their own microgrids either by themselves or with neighbours. “For V2H, we are looking more at a self-reliant home to almost go off-grid. V2G is focused on the concept of the flexible home, where you want to be able to import as much as possible at the cheapest times and where you are very much interacting with the grid,” Goldsmith says.
How smart does the smart-grid interface to consumers need to get? If Nvidia is correct, pretty smart. The company is developing with Utilidata what it calls a smart-grid chip that will be able to run its own artificial intelligence (AI) models, and which is dramatically more powerful than the microcontrollers found in today’s smart meters. Spieler says the kind of chip envisaged by the graphics-processor maker will be able to run Linux, will have several gigabytes of memory, and will operate in an environment similar to that being designed to support autonomous vehicles.
“The goal is an entire ecosystem that’s running a similar operating system across the different layers to build models and applications. You can move the way down to a smart meter for a single house, you can run them at the substation level, or you can put them in the data centre and run them across your entire grid,” Spieler says.
The idea is to make the grid systems much more responsive to changes and make it easier for grid operators, which are already upgrading their systems with sensors and substation-based computers, to determine if things are going wrong on the network.
“Load forecasting is going to be really important,” says Spieler, adding that fine-grained interpretation of sensor readings may open the door to finding anomalies that need a response. “Is a pole starting to sag? Is that a tree hitting a line or just a squirrel running across it? It will help focus maintenance rather than relying on routine maintenance schedules and visual inspection, which are quite costly.”
More computing power inside EV chargers and home energy systems than is in place today seems more likely as the problems outlined by Project Rev are considered. Though existing EV chargers designed for the UK market can offset starting times to prevent thousands of them activating simultaneously, this may not be enough after a power or communications outage. The interactions between the control loops across thousands of devices on a local network could lead to oscillations that lead to an unstable grid if not carefully managed.
Andrew Larkins, CEO, and founder of Sygensys, sees a point at which chargers or the home systems into which they are integrated, in line with the inverters operated by dedicated renewables installations, become more proficient at responding to the demands of the grid and not just their own consumption. Whether they will need to be full Linux machines running AI models is an open question though this is what Nvidia expects to happen.
“Is it overkill for what’s needed? No. I don’t think you can anticipate what the grid’s going to require and the amount of compute that’s going to need to keep it all in line,” Spieler argues. “Anybody that buys today and isn’t thinking about four to five years from now, let alone 10 years, is underestimating what the grid is going to do.”
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.