Grid services from variable-speed heat pumps
Recent experiments at MIT and Purdue have shown that variable-speed heat pumps can provide ancillary services to the power grid. The experiments involved ductless mini-splits, specifically, though similar experiments have been done with variable-speed fans in commercial buildings. The big idea is that by providing these services, HVAC equipment could (a) earn money, and (b) help integrate more wind and solar power into the grid.
The typical target service is frequency regulation, which involves perturbing power consumption second-by-second at the grid operator’s request. Grid operators use frequency regulation to balance fluctuations in load or renewable generation. The power perturbations are fast and average out to zero over time, so they cause only tiny fluctuations in zone temperatures. Zone thermal mass essentially acts as a low-pass filter.
Links to some studies:
https://ieeexplore.ieee.org/abstract/document/7268771
https://www.sciencedirect.com/science/article/pii/S0378778818321571
https://ieeexplore.ieee.org/abstract/document/7001102
My question to you folks: Does this seem like a good idea?
Downsides: (1) The potential revenues are pretty small — on the order of $25 to $50 per heat pump per year. (2) It’s not clear how the added wear and tear from continuously ramping up and down would affect heat pump lifespans and maintenance costs, or how to do it without voiding warranties.
Upsides: (1) Small revenues are still revenues, and the more sophisticated heat pumps already have most of the necessary communication and control capabilities. (2) Experiments show faster, more accurate performance from heat pumps than most power plants provide. (3) Grid operators need more ancillary services in order to integrate more wind and solar power, so prices and revenues are likely to rise.
This question may be a bit far afield for GBA, but I know some folks here have hands-on experience with these units. I’m interested in your thoughts.
GBA Detail Library
A collection of one thousand construction details organized by climate and house part
Replies
The amount of ramping up/down isn't really adding wear & tear- it's nothing like what happens from turning them on/off. From that point of view it's not really a bad idea.
It doesn't really matter that the speed and accuracy of tweaking modulation rates is faster than most power plants- power plants aren't the competition. Smart car chargers, grid aware water heaters, and grid batteries are the competition. While it's possible to use modulating heat pumps for those purposes, voltage & frequency ancillary grid services are probably going to be better served at a lower cost and even better speed & accuracy with smart car chargers, though the heat pumps would probably still be more than "good enough".
Don't expect the revenues for these services to go up over time. One large grid battery can pretty much wipe out the value & pricing of the ancillary services even at fairly substantial variable-output renewables penetration, as the large South Australian Tesla experience has shown:
https://www.utilitydive.com/news/south-australias-grid-service-costs-slashed-90-by-tesla-battery/523436/
https://www.greentechmedia.com/articles/read/has-teslas-big-australian-battery-killed-the-business-case-for-more
Whether it's one big battery, 10,000 aggregated smart car chargers 100,000 mini-splits, the marginal cost of providing those services is extremely small, so expect the competition to be razor keen. One undersung revenue stream of Tesla's fast-charging networks is what they're collecting in the local & regional electricity markets for grid services. A few years ago some Wall Streeters were speculating that Tesla was taking in substantially more from grid services than the margin they were making on car sales.
Aggregating heat pumps for bidding into capacity markets might also be another revenue trickle, with perhaps a wider range of competition.
I don’t see this as being particularly useful, and I’m an EE that works with power systems a lot. Let’s use the southeastern Michigan power grid for example, which has an installed generating capacity of 11,000 MW. Let’s assume your minisplit uses about 1kw when running all out, because it makes the math easy. Your minisplit is 0.00000009% of the generating capacity in this region. Let’s say we have 100,000 participating minisplit owners. The total of all of them is now 0.009% of the generating capacity. That’s with all of them running all out, which can’t go for very long before you get uncomfortable, so we need a capacity factor in reverse, let’s call that 10%, so now we can use a maximum of 0.0009% of the installed generating capacity. That’s not enough to really make any difference, is so small as to be immeasurable at utility scale.
What you’re basically talking about here is dispatchable load. That means load that can be brought online to suck up excess power that doesn’t otherwise have anywhere to go. On an AC power system, excess generation results in a frequency rise. When load exceeds generation, frequency drops. Nominal power frequency in North America is 60Hz, but if you were to measure that frequency with accurate (lab grade) equipment, you’d see that the line frequency is usually ever so slightly under 60Hz. This is on purpose.
Wind generation and solar are not as predictable as conventional generation. This means they can have short-term variations in production that need to be compensated for to maintain system stability on the power grid. Normally this is done with what is known as “spinning reserve”, which are generators that are running but are being operated so that they don’t actually put power into the grid. If more capacity is needed, these generators can be adjusted in a relatively short period of time to provide needed makeup capacity for loss of other generation (a cloudy day for solar, and unexpected shutdown for conventional plants, or a loss of a transmission line). Very short term extra capacity is usually sourced from peaking generators that can be brought online quickly.
Ok, we can see that the grid handles capacity shortfalls without much trouble by bringing additional generation online. Load is different. The electric company can’t magically bring more load online, so excess generation needs to be shut down. This is handled in severe cases by “overfrequency relays” which take generation offline if grid frequency rises too quickly. This is a BIG deal. Large cascading power outages like the northeast blackout of 2003 tend to be large due to frequency related system stability issues.
Currently the only loads that can be quickly brought online are pumped storage plants, which use power to pump water to a higher reservoir to store the energy for later. There are very few of these facilities out there. Think of these as a hydroelectric dam that can operate in reverse too, and then act like a big water battery.
There are several big issues limiting the usefulness of minisplits as adjustable load: the biggest is there is no current control system. No control system means no way to manage the load. Other issues are that the total amount of load is limited because people can only tolerate so much excess heating or cooling in their houses. Lastly, the load is too small to do much, and there isn’t really a need for high speed variability since the power grid is very large. Thing of a massive slowly moving object. It takes a lot to speed it up or slow it own, so it’s relatively immune to small, fast fluctuations.
Dana is right, electric car chargers are more useful here. A typical car charging cycle might be 8 hours at 3kw or more, and it’s stable over that time period. This load can run at night when there is surplus generation available (I have mine configured this way since nighttime electric rates are much lower). Utility scale batteries are also simpler to implement with easier control systems and more stable and predictable loading ability.
Bill
Thanks for the Tesla links, Dana. I'd heard that regulation markets can be a bit thin, but the South Australia experience really drives that point home.
Capacity markets are definitely interesting. I don't think they're an option for demand-side resources in every region, but where they are (PJM is a prime example), the potential revenues seem comparable to those from frequency regulation. Calls for emergency demand response are infrequent, so a heat pump aggregator could make another maybe $50 per heat pump per year for just being available to curtail load in case of an emergency.
Demand response dispatches can last an hour or two, though. Turning heat pumps down or off for that long would definitely cause some temperature drift and discomfort, for which occupants should be compensated. Or maybe drift could be mitigated by pre-charging thermal mass...
>"Demand response dispatches can last an hour or two, though. Turning heat pumps down or off for that long would definitely cause some temperature drift and discomfort, for which occupants should be compensated. "
Exactly. That's why aggregated EV chargers and water heaters are much more suitable for demand response & capacity market than space heating/cooling heat pumps. Bigger loads, not as time-sensitive.
Most capacity markets have time requirements longer than just a couple of hours. Knocking back your modulation levels by 1000 BTU/hr for 2-6 hours isn't usually going to be a s comfort issue, but it's also only ~100 watts of mini-split power.
@Bill, thanks for the thorough reply.
I agree that the scales involved here are daunting. System operators typically require a minimum capacity of 100 kW to enter ancillary service markets. (That's in NYISO, CAISO and PJM; I think it's higher elsewhere.) I ran some simulations that suggest a one-ton heat pump could contribute about 150 W of flexible electrical capacity, on average over the year. So we're talking about aggregating 1,000 mini-splits just to play in the market.
Good point as well about the small energy capacities involved. Energy capacity depends on the allowable temperature swings and the load's thermal mass. Ballpark estimate: 1,000 square feet of conditioned space (including the air and any light/tightly-coupled thermal mass) provides about 1 kWh of heat storage per degree F of allowable temperature swing above or below the setpoint. That's heat storage; the equivalent electrical storage capacity is smaller by a factor of the heat pump's COP.
Long story short: a one-ton mini-split serving 1,000 square feet and allowing temperature swings of +/- 1 F, is (roughly, on average over the year) a 150 W / 250 Wh battery.
Small potatoes compared to an electric car or a resistance water heater!
>”Long story short: a one-ton mini-split serving 1,000 square feet and allowing temperature swings of +/- 1 F, is (roughly, on average over the year) a 150 W / 250 Wh battery.”
That’s about the same as a 12v 20Ah battery, so you have approximately the same energy storage in your house this way as a small computer UPS or about 1/4 of a typical car starting battery. Not very much.
A lot of the clever smart power stuff just isn’t very practical at utility scale. California does tend to be more accommodating of small systems, but it’s due to the regulatory environment there and not anything about the grid. The system operators out there actually have problems because of the mandates to allow some of this stuff.
I really thing in terms of power distribution system evolution, it will make the most sense to allow sheduling of certain large loads (electric car chargers and air conditioning mostly). Car chargers hold the most promise because of their relatively large stored energy. I do not see this stored energy being used for grid support (feeding power back to the grid), but I do see it being useful for load leveling (evening out system load throughout the day).
Real grid scale power storage is ideally handled by pumped storage systems, followed by utility scale battery systems which are typically NOT distributed in very small units. There are a lot of system concerns with distributed small energy sources. I don’t think Tesla’s battery system for grid support really makes much sense. Overall operational efficiency tends to be higher for large centralized systems operating at high capacity factor than for distributed little systems where everything runs at a low average capacity factor.
Bill
Makes sense, Bill.
It's funny that mini-splits' high COPs actually make them less attractive for smart grid stuff. Electrical power and energy capacities are 3-4 times bigger for a resistance heater than they are for a good heat pump serving the same load.
I suppose that tension between efficiency and flexibility exists for any smart electrical load, though.