Clean energy when you need it
Thursday, 15 April 2021
Clean energy advocates think we can use the batteries in an electrified road transport system to balance the supply of intermittent renewable electricity with the demand for electricity, including electrified heat and transport. They are wrong. It can work to a small extent, but it barely touches the sides of the problem.
A well-known problem with low-carbon energy is that its production profile does not match the demand profile. Wind is irregular. Solar is variable (by season, day and weather). Nuclear is more economic run as baseload to spread its enormous capital cost. We refer below to the combination of wind and solar (and similar, e.g. wave and tidal) as “intermittents”, and the combination of these with nuclear as “inflexibles”.
Electricity demand is not well-correlated with these production profiles. Wind is slightly positively-correlated because both its production and current demand are somewhat higher in winter than summer. But on a half-hourly basis, there is no connection between the level of demand and the output from wind. Solar is negatively correlated on an annual basis because of the same effect as wind, but acting in the opposite direction (i.e. solar is mostly summer when demand is lower). Solar’s diurnal pattern is however somewhat correlated with the diurnal demand profile, but far from perfectly. Nuclear’s output is uncorrelated with demand.
That is now. The problem becomes more complicated as heat and transport are electrified in order to decarbonise them and hypothetically to absorb the excess intermittent output that policymakers have decided is required for decarbonisation. If we are to assume below that there are huge numbers of vehicle batteries, then we must also assume that there is an equivalent amount of electricity consumption for transport.
Heat is more complicated, as there is an ongoing battle between electrification and hydrogen. But assuming that we have electrified all the heat is the most generous assumption, as the conversion and storage efficiencies of electrified heat are several times higher than for hydrogen. If your counter to what follows is “ah, but hydrogen”, then what you are really saying is “ah, but the solution is to make the process 5 times less efficient”. Ummmm, no.
After years of trying to play down the significance of the imbalance problem, clean-energy advocates are being forced to think harder about it, as increasing capacity of intermittents causes increasing periods where production is either problematically insufficient or so excessive that generators have to be paid to turn off (“curtailed” or “constrained”).
One popular deus ex machina is to assume that road transport will have been largely electrified, and the capacity of electric-vehicle (EV) batteries can be used to balance electricity supply and demand. The numbers seem so enormous that the tree huggers feel exonerated, as usual, from considering the detail. Surely 35 million electric vehicles will have enough storage capacity to deal with the problem? Well…
There are at least two variants of this solution. The simplistic one is simply to count on the storage capacity of the EVs. A more sophisticated version points out that, if EVs roll out rapidly in the coming decade, by 2040-50, many of the original batteries will need to have been replaced. The old batteries will have diminished capacity, but their cost will have been recovered. They can be put to a second use at low additional cost as supplementary storage. Let’s start with the simple version, as it puts the sophisticated variant into perspective.
Scenario and assumptions
We will assume generously that the average storage capacity of an EV is 100 kWh. So 35 million EVs will have storage capacity of 3.5 TWh. That’s a bit over one percent of current annual electricity demand. But of course the hope is that it charges/discharges frequently, so only a fraction of total demand is required as storage capacity.
On the other hand, that’s current electricity demand. It will be a lot higher when we have electrified heat and transport, even allowing for the large efficiency improvements of (a) heat pumps vs boilers, and (b) electric motors vs internal combustion engines. To be as generous as possible to this scenario, we have taken these efficiency improvements for granted, without worrying about, for instance, whether the building stock can be retrofitted to make it suitable for heat pumps at high efficiency.
We also take a generous/naïve view on the timing of charge/discharge cycles for EV batteries. To the extent the energy is used for transport, we assume it can be done predominantly off-peak (primarily at night, but also in the mid-day dip). To the extent excess battery capacity is used for balancing the network, we assume it can be done as required, without worrying about whether the vehicles are actually plugged in. These are very generous assumptions.
With these assumptions, we arrive at a peak figure for night-time charging of EVs for next-day travel of around 210 GWh/day (this is not the full amount of energy required for road transport daily, but we are also assuming some mid-day re-charging). The remainder of the 3.5 TWh is therefore nominally available for balancing. There may be some standalone (i.e. non-EV) grid-balancing storage capacity, so we (again generously) assume that the full 3.5 TWh is available for balancing services.
We don’t worry about whether the grid operator or electricity supplier (whoever is managing the smart-meter-controlled EV chargers) is actually able to know how much energy each vehicle will need for its journeys the following day. We naively assume they are able to optimise, probably by giving each driver an ability to override the smart meter’s choice. This ignores behavioural issues that would likely reduce the capacity available for balancing.
Electricity demand in an electrified world
These and many other assumptions give us the following figures for electricity demand under this scenario (“Proposed”), and as it was in 2017 (“Base”) for comparison. The 2017 figure may look a little low, but it (a) is end-user demand, not transmission-system demand (i.e. after distribution losses), and (b) allows for conversion efficiencies. This is necessary if we are going to take into account significant improvements in that efficiency. Fortunately, the conversion efficiency of electricity-use (as opposed to the conversion efficiency of electricity-production) is generally very high, so the difference with published statistics is not that significant.
The model allows us to take any of 2016, 2017 or 2018 as a base. That is partly because of the availability of the weather data needed for the model (e.g. estimating hourly heat demand). And partly because that was a sweet spot where there was enough capacity of the inflexible technologies that the output could be used as a basis for the model, but not so much (especially so much embedded) that it distorted some of the figures on which the models rely (e.g. demand). The base year determines only certain profiles whose dependence on the weather (e.g. intermittent output and heat demand) makes it important to use real-world data as a basis, which is amplified by the assumptions in the scenario about installed capacity and usage. The model is not tied to 2016-18 technology characteristics, e.g. it allows for higher efficiencies for new capacity than for the capacity existing at the time.
How do these annual figures look on an hourly basis? 8,760 periods do not visualise clearly, so here are a couple of sample months:
And a couple of sample weeks to decompress further:
The Road element is just the part of the EV battery-charging that will be used to power the vehicles’ journeys. The charging/discharging for balancing services comes into it later.
Once we allow for transmission and distribution losses, the annual transmission-system (“Wholesale”) demand profile looks like:
Again, 8,760 periods don’t compress easily into that space, so here are the sample months:
And the sample weeks:
Peak wholesale demand (around 129 GW) is nearly 3 times higher than at present. We are focused on storage in this article, so we quickly pass by the implicit network reinforcement costs, likely all the way down to most properties.
Inflexible electricity generation
How does that profile compare with inflexible electricity generation? This is a supposedly decarbonised scenario, so we have had to assume a large capacity of inflexibles:
Offshore wind: 60 GW
Onshore wind: 30 GW
Solar PV: 20 GW
Nuclear: 20 GW
Biogas: 1 GW
The annual picture is again a mess, but it does show some important features, such as large gaps below the demand curve, especially in Jan/Dec, and high peaks above the demand curve, especially in Jun-Oct.
Zooming in to the sample months:
And the sample weeks:
This is all before any balancing, whether through storage charging/discharging, or through import/export on the interconnectors. But notice that the periods in winter where inflexible production exceeds demand are so few that there is little opportunity to re-charge storage for months. We could have assumed higher capacity, but then we would have had much too much excess production in summer, and still problems in winter, because a multiple of a small number (e.g. when the wind isn’t blowing and the sun isn’t shining) is still a small number.
Combining these, here is the electricity demand net of inflexible output (negative points are where inflexible output exceeds demand):
Balancing supply and demand
The system needs a lot of balancing. My first run of this scenario assumed complete decarbonisation, i.e. no fossil-fuelled generation. But you can’t get close to balancing this profile with any realistic combination of interconnectors and storage, so I reintroduced 40 GW of gas-fired generation.
We assume that storage will be used as a priority. The interconnectors have second priority, with flows determined by a function of (a) the balance of supply and demand in that period, and (b) the flow in that period in the base scenario, to reflect the reality that our neighbours are not connected solely to take or supply electricity as we require, but also to balance their networks. If our neighbours are as wind and solar-heavy as we are, and it’s a dark, still period in winter, they may not be able to supply us even though we need it. Whatever imbalance remains is served by the gas-fired generation, plus 4 GW of biomass-fired generation and 1.6 GW of hydro, which are also assumed to be dispatchable (another generous assumption, as hydro depends on reserves, and biomass may be hitched to CCS that demands baseload operation).
Note that this means a total capacity of 176.6 GW, way above peak demand in the scenario, let alone the 45 GW (or so) at which the system currently peaks. Apart from finally killing the last remaining coal-fired power stations, we haven’t saved much fossil-fired capacity. We have just (a) massively increased the amount of non-fossil generation, and (b) substantially reduced the periods in which the dispatchable generation operates, which will increase significantly the cost per MWh of that generation. Here is how the technologies contribute in this scenario (“Proposed”) vs 2017 (“Base”):
8,760 periods multiplied by 18 technology layers plus a demand curve is an illegible mess at this resolution. So let’s skip straight to the sample months:
And the sample weeks:
Now we start to get to the crux of the matter. Notice the white space below the Gross demand curve in the week of 22 Jan. These are periods when 3.5 TWh of storage capacity, 10 GW of interconnector capacity and 45 GW of dispatchable electricity still cannot remotely fill the gap between demand and the output of inflexible generators. By design or necessity, these are periods of “demand-side response” to use the euphemism for switching off supplies to users whether they want to or not.
With adequate incentives (e.g. rewarding those users who reduce their demand, which means negative prices for electricity) there will be some voluntary demand-side response. But this is up to 45 GW needing to be cut. That’s around maximum demand by all users combined at the moment.
One reason it’s so high is we have electrified heat, and the weather is cold, while the wind isn’t blowing, the sun isn’t shining, and neither has done for so long that our storage has been discharged. We’ve got 3 days in a row where we need most people to turn off their heating while the weather is particularly cold. Or else turn off all lights, appliances, offices, businesses, etc. and still have to turn down the thermostat. Or tell everyone they have to stay home so there’s no EV charging for travel, and even then still do a lot of the other options.
This isn’t a realistic option politically, but it’s an unavoidable choice when the weather is unhelpful, as it will be every few years, under this scenario (which is what the politicians are currently aiming for). 3.5 TWh of EV battery storage doesn’t begin to scratch the sides of the storage requirements to cope with this scenario.
Electricity storage (mainly batteries)
How is that storage performing? Here is the messy, indistinct annual picture. Notice that the number of peaks and troughs is relatively limited, which will be reflected in the storage costs, which are inversely related to the charge/discharge frequency. This is very expensive storage, even before we start adding more capacity to deal with the gaps.
The red line is the hourly flow in/out of the batteries (negative means charging, positive means discharging). The purple area is the level of charge on the batteries in each period. The model of course assumes that the system cannot be charged above 100% nor below 0%. So if the charge is 100% and the system has excess inflexible output it needs to dump, it will be exported via the interconnectors if possible, and failing that, the output will need to be curtailed. And the logic is similar for 0% charge while the system is calling for balancing supplies.
This is a necessary constraint in the model mirroring an unavoidable constraint in the real world. Many clean-energy protagonists ignore this by dealing simply in aggregate figures and not worrying about whether it actually works on an hour-by-hour basis. You can’t in reality take the excess electricity you need to store in one hour but couldn’t because the batteries were fully charged, and use it in subsequent hours to charge the batteries or supply demand. This is a typical fallacy that aggregate treatments (i.e. dealing in annual or monthly or even daily aggregated figures) conceal. Electricity has to go somewhere immediately, or be curtailed.
Another factor the model has to take into account is the round-trip efficiency. We naively accept the high (c.90%) round-trip efficiencies claimed for batteries. If this were one of the technologies considered more economic for longer-term storage, such as hydrogen or compressed air, the round-trip efficiencies would be lower (much lower in hydrogen’s case) and the problems (other than cost) would be exacerbated.
The more storage we have on the network, the fewer charge/discharge cycles would be made by additional storage capacity. We illustrate this reality with a chart of marginal storage utilisation. If we imagine that the storage is composed of 100 GWh tranches, the chart illustrates how many GWh/year each additional tranche would charge/discharge.
The first tranche would charge around 4 TWh/year, i.e. 40 cycles/year. This is very low compared to the usual ratio of one or two cycles per day required for storage to be economic. But this storage is serving two purposes (the other is being available for long-distance journeys), so perhaps we’ll give this a pass, even though it doesn’t look like an efficient solution.
But look at the decline. By the time we’ve got 1 TWh of storage capacity, we are down to 27 cycles/year. At 2 TWh, we are down to 16 cycles/year. The last 0.5 TWh is hardly used at all, even though we are short of storage. Imagine how much worse it would be if we installed standalone storage to provide additional balancing services. It would be trying to recover its cost on one or two cycles per year.
The problem is that the imbalances are significantly seasonal, not just diurnal. The first tranches can pick up the diurnal imbalances. What’s left is storing excess from summer to be used in extreme periods in winter. It’s not a rational option. And yet, despite assuming that we have installed an irrational amount of storage, it’s also not remotely enough to avoid periods of insufficient supply of up to 45 GW!
The adequacy of intermittent generation plus EV batteries
Putting it all together, this is a chart of the adequacy of this scenario to meet demand.
The blue line is wholesale electricity demand. The green bars are the margin (i.e. excess capacity) when the system is able to meet demand. It is in general around 25 GW. But there are periods (the red bars) where the gap between all the supply options (generation, interconnectors, storage) and demand is as much as 50 GW. We need upto 50 GW of supply that will be used only occasionally and/or upto 50 GW of “demand-side response”. That is very expensive, not only economically, but in terms of welfare and even mortality, as it implies people going cold for several days in the depths of winter.
What if we also used all the used batteries from older EVs? Well, their cycle frequency will be one or two per year, so their cost needs to be nearly zero. But there is an operational problem, because sitting for long periods with either a full charge or zero charge is a killer for batteries, yet that is exactly the purpose for which we want to deploy these. We need around 2 TWh to fill the gap in the above scenario, before we even think about reducing the 40 GW of standby gas-fired generation. There won’t be enough defunct EVs in a century to supply the batteries we need to ditch the gas, even if it were viable and operationally feasible, which it isn’t.
So, no. The batteries in a 100%-electrified road-transport fleet aren’t remotely enough to deal with the problem of balancing inflexible generation against variable demand profiles.
Carbon footprint
And by the way, this has a low carbon footprint, but it’s far from Net Zero. The 40 GW of gas-fired generation is unavoidable because balancing the system without it was ridiculously bad (the red bars in the chart above spread thickly across the year). We can try to decarbonise it with CCS, but CCS not only hurts the efficiency (meaning we need more capacity), but also pushes strongly towards baseload operation to cover its costs. CCS on gas-fired generation used infrequently and intermittently is unlikely to be an attractive proposition to investors.
We also passed over the question of decarbonising forms of transport other than road vehicles (especially air transport). We have not considered decarbonising non-energy greenhouse-gas emissions.
And yet we have bankrupted the country and frozen most of our old people to death with a system that is utterly irrational if the consequences were considered carefully. Instead, they have been dismissed because a government objective once set means that obstacles must be assumed away, or because the government and its advisers (and supporters of this approach) are so naïve that they do not understand that the system actually works on a minute-to-minute basis, not an annual-aggregate basis.
Public choice and the knowledge problem (again)
This is the reality of central planning. It overestimates its knowledge, and serves the interests of those who do the planning, and those who can extract rent by influencing the planning. In this case, the mechanisms needed to force the market to deliver an irrational system provide huge amounts of rent for the large energy companies who install and operate the systems and who are “de-risked” by the support mechanisms. The politicians claim the credit for their environmental compassion and strategic industrial vision (strangely for a group who between them have little commercial experience), whilst denying the blame for the impact on costs and lives (the rent-seeking corporations have to suck up the blame for their “profiteering” so the politicians will put the mechanisms in place to make that possible). The civil servants get large empires controlling the managed markets that are necessary because no free market would do such a thing. Government advisers (academics and consultants) get the contracts to decide what the market should be forced to deploy, and the kudos when the market delivers what it has been forced to deliver, or the additional contracts to work out what went wrong and how to fix it if the market doesn’t deliver what it is supposed to deliver. The rest of us get to pick up the bill. But we voted for it, so...
This is not fundamentally different to any managed economy that permits private ownership and profits so long as they are engaged in support of the government’s objectives, whether Lenin’s NEP or the Nazis’ relationship with big business. The greatest sleight-of-hand in modern politics is a bloated, overmighty state pretending that it is market-friendly because it allows large corporations to profit from delivering its objectives. That is authoritarianism, not voluntary exchange.