Earth Notes: On Distributed Grid Support From Microgeneration (2007)Updated 2021-02-28 11:36 GMT.
Larger organisations with substantial 'instant-on' battery storage may already participate in NG schemes, using (for example) their diesel backup generators. These may include telephone exchanges and Internet hosting datacentres. They might also be prepared to add their 'instant' resources to the pool I describe, eg while their battery stores are at over 90% full.
To this end I have exchanged a number of emails with NG (see below). I am grateful for the courtesy with which my enquiries have been handled.
It might give a 'warm feeling inside' for a householder to be able to help in a small but concrete way with energy security. It might also be possible to find a way to do this well enough to be worth a regular income to the owner of a grid-tied system. This on top of any payment for energy generation and exports.
None of this is likely to make installation of a solar PV system (say) a moneyspinner. It may, however, provide some concrete help in frequency/load support. It may further incentivise behaviour useful to the grid and to energy efficiency. Thus it may also possibly towards the UK's climate-change goals.
The core element of this is injecting extra energy when needed into the grid. That is, when frequency starts to fall too far due to excess load. Or withholding available injection of energy, such as from PV in high sun, when frequency starts to rise too far due to excess supply.
This is similar to time-of-day (time-of-use, ToU) metering in externalising costs. The end user is charged something strongly related to the instantaneous wholesale cost of each unit that they consume. A microgeneration system with local storage might effectively 'island' the premises and avoid using grid power at all during peak demand where carbon-intensity and grid strain is at its highest. And indeed at other times where imported unit cost is exceptionally high.
My initial enquiry to NG in November 2007 was as follows:
Do you have any contracted fast/standing reserve as small as a [G83/1] (11kW/3-phase) connected via an urban DNO [Distribution Network Operator]?
I ask because I note that it might be capable of meeting your 25MW/min requirement (ie one-cycle start on demand in principle) from battery backup, and would only need a relatively small (extension) battery bank to cover 11kW for a 30-minute period.
I am currently outlining a small solar PV system that could in principle provide some such grid fast reserve and/or peak load support, including at winter peaks, and though rather fiddling in terms of output, might be useful to have injected directly into the urban network.
Is 11kW below your noise threshold, or is it worth attempting to take this further?
Two important issues here are:
- The small size (low power) of the potential generation (11kW) compared to the typical outage that NG might have to deal with (~660MW). Also the magnitude of the total load on the system at winter peak (~60GW).
- The fact that the generation is within the local distribution network and not to the grid directly. The could be an advantage in that the local injection might effectively be boosted by the distribution network loss displaced (~7%). Or it could be a disadvantage to the NG being too decoupled from them.
NG's response in mid-December was to the effect that there are various minimum requirements to provide "Fast Reserve" service. The plant:
- must be able to start delivery within 2 minutes of instruction
- must have a run-up or run-down rate of 25MW per minute
- must be capable of sustaining output for at least 15 minutes
- must be instructable in blocks of at least 50MW
and there were some clear problems with my suggestion. Such as the small (unaggregated) size of a G83/1 inverter cf the 50MW minimum, and power delivery from a system deep within the DNO network.
NG suggested that my scheme might be more suited to providing Short Term Operating Reserve (STOR). The minimum criteria for this service are not as dynamic as those for Fast Reserve. STOR plant:
- must be able to start delivery within four hours or less of instruction
- must have the capability to sustain output for at least two hours
- must be instructable in blocks of at least 3MW
- must have a recovery period less than 20 hours
- and be available at least 3 times per week
Here NG is trying to think bigger in terms of total energy injection. The battery-backed G83/1 grid-tie system can't easily achieve that in terms of permitted power export limits and battery-bank size. What it can possibly achieve instead is nimbleness. Such as possible single-cycle response and/or response after a small random number of out-of-spec cycles. Maybe with some exponential backoff in serious cases to help avoid introducing any instability itself. These are well-known techniques from computing and mobile telephony to make large swarms of small actors behave well as a group.
Thus I responded:
Clearly the small system I suggested fails for Fast Reserve requirements on the [last] of the requirements, ie instructable in blocks of at least 50MW, by more than 3 orders of magnitude!
In the US the is some talk of 3rd-party aggregators that would coordinate tens of thousands of households' grid-tie outputs.
I assume that there is no such aggregation scheme active or on the drawing board in the UK that you are aware of? (Given that only hundreds of domestic-scale export meters are currently installed...)
In fact the STOR would still need lots of aggregation to be interesting to you by the looks of things, and the longer run time is likely to be more problematic in terms of capital cost of battery storage (else you'd have a warehouse or two full of them contracted to you already I imagine).
I was thinking of an entirely-automated distributed system which automatically kicked in when frequency fell below (say) your -0.2Hz threshold, and equally would stop exporting to the grid overriding other considerations when a (say) +0.5Hz threshold was passed. There might then be a simple meter in place to indicate how often/long the installation had been able to export extra power when needed, and stopped exporting when instructed, and NG (or the aggregator) would pay on the basis of that. (On a larger scale and longer-term, possibly as an addendum/extension to the G83/1 spec for example.)
This would need no central instruction at all, but would mirror your automatic load-shedding mechanisms for large commercial concerns, as I understand them.
NG's response was to the effect that its specifications for Balancing Services are based upon MW sizes that are large enough to have a visible benefit on the transmission system and that are economic to dispatch, meter and settle. Traditionally this has meant big blocks of generation and demand-side regulation. And though NG is seeing more participation from aggregators to provide STOR services, those services still need to be delivered to NG in multi-MW chunks to manage efficiently. NG doesn't do retail, but under its license obligations to operate economically and efficiently would consider any economic alternative to existing providers that can meet technical/service requirements. Automatically-triggered systems would still need to report availability in real time to be of use to NG.
(NG directed me to its (interesting) document on Operator Incentives http://www.nationalgrid.com/uk/Gas/soincentives/consultation/ which it invited me to look at (half way down page 43) and respond to. The section concerned seems to be about NG making money available for development, which is not important to me at this stage. I'm trying to see if there would be a positive and valuable role for microgeneration to play soon in grid stability, with a possible new revenue stream over and above pure energy sales and/or ROCs.)
I'd argue that NG is looking at the distribution problem from the wrong angle. Virtually every current office network is based on the 'Ethernet' design with no central control and many, many participants. WiFi LANs and mobile phones do have a stronger central coordinating system, but would again be non-functional without significant distributed intelligence/control. Although I can see that it is much easier for the NG to control large lumps of generation on demand, if we move towards something like that envisaged in the [archive] Home Truths report where there is in effect no baseload, and households are net power exporters, then the NG and its successors are going to end up dealing with something like my suggestion anyway, I believe.
I fully accept that some cases cannot be completely covered by, for example, the distributed system outlined above triggered by line frequency alone. Since the UK grid is synchronous then its frequency is uniform everywhere. But there may be transmission constraints due to outages or upgrades, etc. So a lack of power in Land's End is not helped by injection of more energy at John O'Groats and vice versa. While the distributed system remains small this effect is not catastrophic. But it needs to be dealt with by improved 'intelligence' if and when its contribution grows relative to centrally-dispatched balancing support.
(Note that a grid-tie inverter's response to deviations from nominal/centre frequency can be deliberately made asymmetric. The inverter can continue to power the grid for a much bigger drop from nominal than rise from nominal (frequency primarily, but maybe voltage too). Then the inverter/microgenerator is effectively providing a type of 'reverse dynamic demand' support to the grid.)