Written evidence from Stephen Broderick (EVD0062)
Basis of Opinion: This is my doctoral research area at Southampton University. The work is as yet incomplete (I am in my final year). To research the topic, I have developed a system able to simulate, study and manage networks with EVs undertaking various trip duties. Further, the below is informed opinion based on observing / modelling likely UK situations but not proven in practice.
Comments are the authors own and relate to home charging of EVs on existing Low Voltage (LV mains) distribution networks.
Summary: The UK Distribution network "as is" is adequate for immediate needs, but will be substantively overtaxed by unconstrained EV home charging (by up to 7:1), for EVs draw c. 7 kW for hours. These issues follow uptake of EVs i.e. minimal at first then overwhelming as years pass.
Consequences will depend on local circumstances, but have potential to include:
• power cuts (overloads of supply equipment => power equipment "blowing fuses")
• brown-outs (loss of sufficient voltage) potentially causing:
◦ home appliance damage and
◦ household fires
• potentially, a move to restrict EVs to (say) 1 in 7 homes.
Two general methods are seen to alleviate these situations:
- reinforcement of the networks (asset replacement) - expensive, slow, disruptive
- use of a control system capable of managing EV home charging.
Further, the EVs would need to obey the issued commands; this is not assured.
Note that the ICT / SG method does not provide a complete solution, but is expected to defer major costs for decades.
Background information to assist the reader:
- A brief summary of technical terms is appended.
- The UK “electricity grid” is a set of millions of connected components, individually selected to affordably perform a given goal. These parts have been installed in an as-needed / piecemeal manner since c. 1900. Many parts date from the 1950 / 1960's
- The grid can be broken into 3 main sections:
- Supply (the making of electricity)
- Transmission (sending electricity at high voltage around the country e.g. National Grid)
- Distribution (of power at medium and low voltage to customers e.g. "mains" 230 V)
- Many papers have been published concerning EVs and "the grid"; however most relate to US networks – similar to the UK only at higher levels. The lower Distribution level is different, so to meet the challenges of different situations.
• US: adequate (occasionally challenged) Generation and Transmission. Strong Distribution which can provide 8 – 14 kW to each home simultaneously, the major load being Air Conditioning.
• The US Distribution system typically uses many “near-home” local transformers each supplying 1 – 4 houses;
• UK: adequate Generation, strong Transmission. Adequate Distribution (for present loads) able to provide 1 – 2 kW to each home simultaneously.
• The UK Distribution system typically uses a small number of “substations” with a transformer, each supplying 1 – 100's of houses.
Reiterating the US / UK difference in Distribution capability:
◦ US: 8 – 14 kW per home, able to sustain peak loads for long periods vs.
◦ UK: 1 – 2 kW per home on average, able to support occasional higher loads due to averaging over many customers
Most published papers originate from the US so silently assume the US model. However the respective Distribution networks have quite different characteristics.
* * * Overseas studies and experience-based advice may not relate to the UK.
Some Numbers:
a) there are about 250,000 Distribution networks in the UK, each of individual nature;
b) the design of these has historically been guided by a measure called "ADMD" which has been set variously to c. 1 to 2 kW per supplied home. This is a statistical measure and assumes customers take random loads at random times;
c) to drive an "average" day's distance (c. 27 mpd) an EV consumes c. 9 kWh at the wheel;
d) (at the time of writing) batteries loose c. 8% on charging and the same on discharging; the inverter electronics lose a similar percentage;
e) the daily average EV power draw (at the home charging point) is then consumed power plus losses i.e.
9 * 1.08 * 1.08 * 1.08 = 11.3 kWh
Other aspects cause losses; 12 kWh is a reasonable "EV supply average daily demand".
f) driving distances are, in general, dependant on location / nearness to a city. RAC studies suggest the following mileage ratios:
City : Urban : Rural of 1 : 1.4 : 2 (i.e. country dwellers drive twice as far as those in the city)
g) a modern EV home charger draws 7 kW.
A 100 Home Illustration
A 100 home development built in 2017 has an LV distribution network fitted to supply 150 kW simultaneously. An overload of c. 50% is possible for up to 8 hours (following this a period of cooling / low load is necessary). This network includes:
• substation (with transformer and switching)
• in-road cabling (3-phase 230 V per phase)
which has an asset value of c. £ 30 - 45 k.
If 100 EVs arrive in the evening and start to charge, the peak load is 700 kW and the distribution assets go into immediate substantive overload.
However the EVs require (on average) 100 * 12 kWh => 1,200 kWh of energy, which if supplied in a staggered manner over 10 hours is 120 kW continuous load thus doable.
This suggests that a local rationing or management system (able to pace demand intelligently) would help. A simplified version of this was successfully trialled in 2015 (My Electric Avenue).
Yet this has ignored the usual demand of the households; in winter they will need power for home use. In extremis, it may be necessary to upgrade (reinforce) local substations and cables.
Such reinforcement includes:
• replacing the transformer
• digging up the road and relaying cables
• a spend of c. £45 k per 100 served homes i.e.
25 million * £ 450 = £11.25 billion (approx assets costs; a "broad-brush" estimate)
Which, after manpower costs plus profit is added may be perhaps x 2 or x3 as much.
Note that the politics to this falls into three sections:
- the money - who pays?
- the inconvenience, primarily digging up the roads to relay cables (especially in cities)
- the manpower - with the best will in the world, this is a project which may take a decade or more - with present manpower. To achieve this faster requires more hands.
Hence, a new generation of electrical engineers and technicians are needed.
The initial threat though is simultaneous arrival and charging; even with 1 in 7 of homes having an EV the system is at full capacity in the early evening.
NOTE This ignores home-heating by Heat Pumps (HP) scheduled from c. 2040 on as part of the UK's CO2 minimisation initiative.
** HP alone impose more load than EVs; immediate reinforcement will be necessary **
Select Committee Questions
Q1: How will increased uptake of electric vehicles, to meet the Government's 2040 target to end the sale of new diesel and petrol cars, affect the electricity grid?
A1.1: SRB defers re Generation and Transmission. Concerning Distribution:
A1.2: For present LV networks: With EVs at 100% penetration worst-cases will happen, as there are so many individual Distribution networks in the UK that the number of daily "dice-throws" occurring mean that even low-probability events can be expected. These include "random" power-outages and damage (worst-case: fires), likely initially seen in:
• the most affluent areas (faster EV uptake is expected there) and
• those areas with the highest driven distances i.e. rural.
These are low-probability events but have hight impact on those effected.
Q2: What action is needed to manage impacts, and to make the most of opportunities afforded by vehicle-to-grid technologies?
A2.1: A workable, locally informed, secure Smart Grid is likely needed, so to allow a form of computerised management. Today, this does not exist. The Internet is indeed present, but ultimately is so insecure as to be specifically avoided.
Note the culture gap which suddenly yawns: Power engineering is paced over decades to centuries, not short consumer cycles typical of IT or electronics companies. IT security is a perpetual arms race; what was secure in one decade may be easily broached the next.
Any Smart Grid used to manage the nation's power needs to meet military security standards. The reason for this is that society will likely use electricity for centuries. Unless there is a guarantee of peace throughout such extended periods, exceptional security is necessary otherwise the society is readily defeatable in times of war.
This is highlighted by the Ukraine's experience, where power systems have been remotely hijacked. Power companies have seen this and are wary of the Internet.
Concerning EV charging load. The simplest route forward is to use an ICT approach, so to stagger EV charging as suggested in the earlier 100 Home Illustration. If so operated, the majority of the present networks will be adequate for perhaps several decades.
Each local network and substation needs be incorporated into such a scheme, so that decisions can be localised / based on knowledge of local conditions. Today, there is no requirement to monitor substation loadings; real-time load monitoring is rare.
Access to such data and the accompanying controls over EVs can be abused by malicious parties. For instance, all the EVs in a local network could be told to charge or discharge simultaneously; this would certainly provoke the stated problems. Note that control over monitoring data can achieve the same thing; if the monitoring data were to be manipulated to claim there was great spare capacity, all EVs might simultaneously choose to charge (or conversely - to never charge). Direct control over the EVs is not needed to cause havoc.
V2G
The opportunities afforded by vehicle-to-grid (running the EV charger backwards, technically straightforward) are, in the UK context, likely minimal (further amplified on in the Appendices).
Why is this, given the excitement and prolific research into this area?
Here I discuss:
A) the hope of V2G,
B) the consequences of these with focus on the UK context and
C) an alternative which side-steps the problems arising.
A) The hope of V2G
V2G may allow a management party (known as an "Aggregator") to operate massed groups of EVs to offer network support services. The suggested services fall into two broad categories:
- network frequency control, and
- bulk (peak) power support.
Concerning (i): If a network occurrence (e.g. loss of a power station) causes electricity supply to drop below demand, energy flows from the mechanical inertia of (other power-stations) rotating generators; these physically slow and system frequency drops. In order to re-establish balance and restore frequency there are two methods:
• acquire new sources of power, or
• shed (dispose) of load.
Concerning (ii) bulk (peak) power support: Demand varies by c. 80% over the day, with (in the UK) a demand peak in the afternoon. Keeping "spare power stations" ready to supply this peak is expensive; they are only used occasionally / stand idle for months - plus the act of being turned on / off for short periods induces excess wear.
Given a large group of parked EVs (say a million) each with a 7 kW inverter - then a 7 GW power station, operable for minutes to hours, is available. Contrast this to the proposed 3.2 GW Hinkley Point C nuclear power stations. EVs will likely become the biggest power resource on the Network.
This prospect is of particular interest in the US, where network support services command fees so earn income. This has been extensively researched and is, technically, completely viable.
B) The consequences of these with focus on the UK
Note that a V2G EV dispatching (sending back) power to the grid will need to subsequently recharge. This may be deferred but is likely necessary before the EV departs its parking position. Also note that the following assessment is riddled with assumptions about the demands / needs, the battery technology and susceptibility to damage / rapid ageing.
As battery technology matures, limitations and capabilities will likely change.
The extra charge movement implicit in V2G operation:
- may age the battery prematurely
- may restrict use of the car, for "readiness time" is extended by the time for further recharging
a) Will V2G harm the EV? Possibly.
The vehicle battery is doing more work, so is likely to age more rapidly. However recent studies have suggested that small amounts of V2G may aid rather than age the battery. Frequency support duties are likely less damaging than peak support duties.
The net result is: Damage hence costs are possible and will effect some EVs.
An example:
The Nissan Leaf battery c. 2015 was good for c. 1,000 charge / discharge cycles and held 24 kWh. Battery replacement costs change rapidly, but have been cited in the region of £5,000 to £8,000.
From this we can see that one charge / discharge cycle incurs a capital loss of c. £5 and the cost of the electricity held is c. 24 * 0.15 = £3.60. Immediately, any income on V2G use of 24 kWh must better £5 of damage and £3.60 of lost electricity. This particular EV is unlikely to profit from V2G.
b) will V2G ever impact the use of the EV as a car (i.e. drain the battery so there is insufficient charge for use)? If mis-managed, Yes.
Here, mis-management is likely to come in the form of the EV not knowing the needs of a family for the use of the car - information which the family themselves may not know.
This could go badly wrong. A car left parked with charge for 200 miles - may be found much reduced later, especially if the car is needed urgently in emergency. Expect headlines along the lines of "V2G stole my charge - I couldn't get to hospital and my baby died" at some point.
c) is there a market able to fund V2G, in the UK? Some (to start), weakening to No.
Information on this issue is available from the National Grid, who operate a reverse action in order to secure their needs. The conclusion is "No" because:
- the UK has limited needs, hence limited funds seek to trade, and
- oversupply is expected from up to 20 million possible supply EVs, competing for a limited market causing the price to slump - thus disappearance of the income stream.
d) is use of V2G without cost? No.
If an EV experiences a V2G discharge then that charge will need be returned. However the Aggregator, having just spent some charge, needs replace the charge "for free". Today's domestic billing and metering systems are not aware of this and will bill the home twice.
This implies a necessary restructuring of retailer billing and metering capabilities, along with data flowing about who really had to pay for the replacement charge. This may get complex.
Note that due to losses V2G is a "zero-sum-plus" proposition; the losses impose penalties.
Example: to dispatch then replace 10 kWh:
• to deliver 10 kWh to socket via inverter: battery depletion is 10 * 1.08 * 1.08 = 11.16 kWh
• to replenish 11.16 kWh needs: 11.16 * 1.08 * 1.08 = 13.02 kWh
Thus 13.02 kWh must be bought to replace the consumed 10 kWh.
e) Is there any risk to the LV networks specifically from V2G? Yes.
(Due to the large number of dice-throws, the situations described are likely to happen)
Clearly, if a remote aggregator is unaware of local constraints, they may provoke issues such as:
- co-ordinated activity breaks the network ADMD design-assumption ("loads are random")
- being unaware of local capacities and instantaneous situations, damaging / overloading commands can be issued
- random location of EVs on transformer phases may result in great unbalances including
- the rare but possible situation in which maximum V2G happens on one phase, whilst there are extreme loads on other phases. This can cause spontaneous extreme heating and transformer fires and is mentioned as this has occurred elsewhere.
Multiple Aggregators bring further issues: they will need to know of each other's activities, else the situation may arise where two EVs standing next to each other are controlled to opposite ends. Aggregator A commands "charge" whilst Aggregator B commands "V2G". The two could sit trading power back and forth; losses dominate, costs are incurred and there is no net result (other than releasing further heat and CO2, to no end). This is further complicated by EVs tendency to move about i.e. may not be where the Aggregator thinks they are.
However there is a simple method which can address all these concerns.
An Alternative to V2G
Consider the case of a large set of V2G capable EVs. Typically each will charge on average 2 hours a day; with an ICT controlled timing system charging is staggered into discrete slots i.e. at any instant there will be a set of EVs charging, but not all.
Telling one charging EV to V2G dispatch will result in a "turn-around" of 14 kW (from 7 kW charging to 7 kW dispatch). Yet the same effect can be gained by telling two EVs to pause charging. This is no hardship for the selected vehicles (just a change of when they charge).
Thus, the method is to suspend load for a period, rather than command V2G dispatch. In situations where there is load to discard this method offer the same benefits as V2G, without all of the above concerns re battery ageing, loss of wanted charge, billing issues, technical complexities etc.
Q3: How do charging infrastructure requirements differ for alternative types of vehicle, journey, and user (including fleets)?
Ultimately the use (driving patterns) of EVs determines the infrastructure needs.
A system with EVs plugged in - but which make no trips - is not electrically challenged. The more the EVs are used, the more challenging a load they offer. Thus:
- location is important as that effects daily EV use (see prior City : Urban : Rural use pattern)
- usage effects load demand; an EV taxi rank with charging will have more demand than a commuter car-park with charging.
Motorway recharging stations need to cater for a very wide range of vehicles and loads and is a specialist area.
Q4: How should new infrastructure for electric vehicles and associated grid reinforcements be sustainably funded?
I have no expertise base here but suspect that Central Government will need to construct a funding vehicle for this; for them the monies (likely multiple low 10's £ billions) are manageable, indeed small if spread over decades - but such amounts are crushing for others.
SRB's Suggested Solution
This is a system arising from my doctoral work and is a working simulation called the "Feeder Phase Balancer" or "FPB", developed whilst working at Southampton University.
This work is funded by the UK EPSRC in conjunction with SSE and has been helped by inputs from power engineering professionals. Example outputs are in the Appendices.
Simplified FPB Method:
1. Bulk EV control by remote Aggregators as presently conceived is removed.
2. The Smart Grid is represented by a local ICT solution, based about a controller located at the substation. ICT connects the substation local controller to EVs on the supplied network, perhaps by WiFi or Ethernet. This isolates each group into a local cell and allows improved security (particularly if the local controller is hard-wired to the EV charging points).
3. the system can operate stand-alone or under remote command in one degree of freedom only: Maximum load drawable limit (power engineers: this is an automated DR system).
4. the controller can measure local transformer loads, and
5. can establish a control connection to most (not all) EVs as they arrive
6. the controllable EVs can include load-managable types and load-managable types with V2G capability
7. each EV proposes a charging schedule; the local controller can see all of these and "time-slices" each EV so they operate sympathetically together
8. unexpected loads and events occur; the controller sees these as changed loads, so will cause EVs to suspend or limit charging or to provide V2G.
The net results are:
- for other than extreme situations, the system is a great improvement on "doing nothing"
- using such a system, most networks can likely have reinforcement deferred for decades. Short-term overloads may still occur (but appear to be within the capabilities of the assets)
- the method is much more affordable vs. reinforcements
- as the local DNO is in charge of this, there are no Data Protection issues (data is local and held by one company, not being sent about or shared amongst many)
- the DNO can mimic an Aggregator, in so far as the DNO controls the "Maximum load / DR setpoint" which may be adjusted up or down as needed. Response may be fast thus able to offer network services i.e. Profits for network services now flow to DNOs
- there is minimal impact on EV battery life from V2G duties; these essentially disappear
- there are no billing complications / need to revise retail billing systems
- there is no fear of V2G induced "charge loss / stealing"
- there is no need for a secure UK-wide Smart Grid on Day One.
Study of the Southampton FPB system is throwing up interesting emergent behaviour and observations:
• although V2G is available for use, the FPB system by preference operates by load reduction. V2G activity markers are very infrequent (other than in the most heavily stressed situations)
• EVs which charge away from home offer substantive benefit to the home LV networks
• the most significant factor is local fleet trip distance
• closely followed by trip timing impacts i.e. unfortunate timings of arrivals / departures.
In general:
• Distance from a metropolis sets driven trip distances
• Trip distances set total kWh demand
• Trip timing sets instantaneous kW peaks.
Further note on Aggregation:
The FPB enables the ability to charge EVs affordably at home. Although not intending to be disruptive, by managing total demand the FPB defeats the work of Aggregators:
That a remote Aggregator desires to "turn down" load is seen as the release of local capacity. The FPB responds by allowing extra charging by other EVs. Thus, no net change.
And vice versa; that which a remote Aggregator "turns up" the FPB responds to by - turning something else down. Again, no net change.
Appendices:
1: A 100 home LV network in Winter with no EVs. Only household demand is shown. Timescale is 10 divisions per hour over a week. This is live load data from measurements made in the London area. The measured load ADMD is 1.5 kW and the network is "Weak" i.e. installed assets designed for continuous 55 kW per phase. Peak load is just over 40 kW
Figure 1: kW load per phase on a monitored Thames Valley network. Losses (green) are under 5%.
2: As (1) plus a dumb EV (various sizes) at each house; average 38 mpd trips. No FPB control. Two plots are shown: the simulations are identical other than a randomiser controlling trip return time.
Figure 2A: As Fig. 1 plus a dumb EV per house (av. travel 38 mpd). Randomiser seed is 101. Losses have approximately doubled and just exceed 10%
Figure 2B: Repeat simulation as in Fig. 2, with return randomiser seed of 120 (travel distances and departure times remain the same; times of arrival home change). Travel habits impact loads. Note change in kW scale; peak load now just exceeds 100 kW
During the simulation for Fig. 2B, the FPB system found 103 "transformer distressed" warnings (i.e. 10.3 hours in the week operating beyond allowable "emergency overrating") and 312 feeder overcurrent warnings (i.e. 31.2 hours in the week).
3: As per (1) with EVs: 20% dumb, 40% controllable charging, 40% controllable + V2G ability. FPB manages the latter two groups. The load control level is 49 kW per phase. FPB preference is to "turn charging down" over V2G (available as a last resort). Return home randomiser seed is 101
Figure 3A: FPB managing 80 EVs. Return randomiser seed is 101.
Figure 3B: FPB managing 80 EVs. Return randomiser seed is 120.
Here, the simulation of (2B) is repeated with the FPB. There were no transformer or feeder cable distress signals. EVs charged some 43,347.9 kWh in the week yet only dispatched 1.7 kWh via V2G.
4. Power Fails - how do they effect the system?
The simulation of Fig. 3B (seed 120, daily trips average 38 miles) was repeated with a Power Fail from 7pm to 4am, applied to the same FPB controlled EV mix:
Figure 4: Overnight Power Fail (7pm - 4am) Return randomiser seed is 120, average daily trips 38 miles
Other than increased load in the periods after the power-fail, Fig. 4 is very similar to Fig. 3B.
Software used: SRB written "FPB" system for managing EVs as if by a local controller; OpenDSS as a network solver (proven modelling capabilities, detects overcurrents, out-of-voltages; generates losses data).
Comments to Car Manufacturers:
Please:
- make ICT very secure. Remotely controllable / variable charging rates are needed.
- if an EV knows it will stand uncontrolled for a long period, cause it to use an adequate low charging rate, in preference to a high charging rate. This eases LV system impact
- design specific inverters for each market. Multi-market inverters may not cut-out on sensing low volts. Charging should always cease if V drops below 216 V in the EU (NB some multi-market inverters designed for 110 V operation may not do this, giving rise to LV network "voltage drag-down".)
Terminology (somewhat simplified)
ADMD / After Diversity Maximum Demand: Statistical averaging of random habits of many people
Amp, A (aka: Current): the flow-rate / volume of flow of electricity
charge: The amount of held electrical energy e.g. by a battery. Can be expressed as kWh or SOC
losses: a physical dissipation of energy (as heat) caused by the act of sending electricity via wires. Loss is tied to:
• the square of amount of current flowing (so doubling power quadruples losses), and
• the material used to make wires and transformers. Cheaper materials tend to be more lossy
kWh: a unit of energy being 1,000 Watts for an hour e.g. heat from a 1 bar electric fire for an hour
kVA: a term used when specifying transformers. For our purposes comparable to kW
SOC: State of Charge: a percentage of battery rating. 33% SOC implies the battery is "1/3rd full"
Volt, V, kV: a measure of the pressure of electricity
Voltage drag-down: a possible characteristic of inverters. EV inverters are constant power devices; thus they respond to supply voltage drop by increasing current loading (so keeping the product fixed). However many inverters doing this in unison increase network losses, causing further voltage drops etc. A destructive cycle forms and local voltage collapses. Not thought to have yet occurred in the field.
Watt, kW: Units of power. 1 kW is approximately 1 1/3rd horsepower. Power = Volts * Current
November 2017