Getting to zero CO2 in a Wisconsin “passive solar” house with radiant heat
Hello,
I would appreciate thoughts and advice concerning how to reduce fossil fuel heating for a rural house. I’ll describe it, the current problems, and the assets. I am fine with spending a bit of money getting to ~zero CO2, but don’t want to break the bank or engage in massive reconstruction. I’m new to all this, so apologies for being a bit long-winded and my general lack of expertise.
This fall I bought a so-called passive solar home in rural Wisconsin, as a vacation home (to be used mostly in the summer). First built in 1983, with additions and renovations since then, it’s an earth-bermed southern-facing house, with a huge amount of glazing on the upper floor. It’s about 1900 sq in total. When it’s sunny in winter, the indoor temperature can easily shoot up from low 60s in the morning to 78F by mid-afternoon. It’s cozy and on a beautiful property. I love it, but the reality is it’s not as passive or solar as it needs to be.
Problems: there’s little thermal storage in direct view of the glazing; the sun shines on a window seat covered in white tile (…) and hard wood floors. Additionally, being in climate zone 6A, sunny days in the winter are not guaranteed, the outdoor-indoor temperature differential can be 80F degrees. The insulation is ok, but not quite as much as you’d want; the previous owner told me the roof insulation (new, cellulose) is at R-35, and some joists are visible. Even on sunny days, the house quickly cools after sunset. The windows are double-glazed, but still quite cold after dark. What’s apparent is that the previous owner relied on the wood stove to heat the house (which does the job admirably). That’s not an option when we’re not there.
Assets: 1) newer, maybe 6 years-old?, insulated concrete slab floor on lower levels with hydronic radiant heat, powered by a Lochinvar Cadet CDNo40 tankless LP boiler (http://www.lochinvar.com/_linefiles/CDN-04.pdf), with output at 9–37 MBH; 2) one mini-split AC/heat pump, several years old, and I couldn’t tell you the make/model/capacity, controlled with a remote and lowest heat setting of 63F. 3) a 7 KW PV array (separately mounted), which averaged 15 KWH/day in February (a 3 KWH/day surplus over that time; I was at the house for a week in that timeframe).
With no one home, the primary heat source has been the radiant heat, and I think usage has averaged out to be ~20 MBH (when it runs, it runs full-tilt), based on LP tank fuel charges. I use Nest thermostats to control the two zones and have each on a strict schedule, increasing the heat to 62F at 6:30pm and then going down to 57F in the middle of the night (midnight for the smaller enclosed zone, 2AM for the larger zone that heats the majority of the house). My not-particularly-well-considered idea here is that I want the slab cold in the morning, in case it’s a sunny day. If it is a sunny day, catching the temperature drop at 62F in the evening should be perfect. I’ve had this schedule in place for a few weeks now and it’s been better than simply having the heat set to 60; if it’s cold and not sunny, the Nest will start preheating the primary zone in the afternoon and run it until the 57F dropoff. When it’s been sunny, then the heat might not kick on until 7pm or 8pm. The thermostat is on the lower level, so there’s several degrees of difference between it and the upper level in the afternoon, with less difference in the dead of night.
The output water temperature for the boiler had been set to 145F. I’ve dialed this down to 115F. Hopefully I caught this in time to keep the PEX from being harmed…
The contract with the electric utility (Alliant) is for parallel generation. They buy it from me with a charge for a second meter. Unfortunately, the contract is up soon, so they’ll no longer pay me $.25/KWH; it’s not clear whether they’ll drop me to net metering or to a price below what they charge me (currently $.126 KWH, including an optional renewable energy surcharge). I can continue to sell to them, or simply rewire to offset usage and avoid the $11–$12/mo charge for the second meter.
Getting to zero CO2 heating will be tough, but these are my ideas for making a drastic reduction. I should note I’m a software engineer and have much of the summer off work this year, so I can make the home “smart” easily enough:
1. Put cellular shades on the windows, and jury-rig some motors to close them in the evenings automatically. I think they are presently the biggest source of heat loss.
2. Increase the insulation in the attic, and pay attention to sealing gaps around vent stacks, etc. (I haven’t costed this out yet).
3. Figure out how to control the heat pump, so it turns on in the early evening. I am not certain what the operating outdoor temperature range is for the heat pump and doubt it can handle the heating load all by itself. The January cold spell went as low as -25F for a few days this year, though that’s abnormal. The minisplit is on the upper level, so it also can’t easily get heat to the lower levels.
4. Get a water heater tank with an internal heat exchanger and run the main input line to the boiler through the heat exchanger to pre-heat the water. I could either buy and install solar collectors to heat the water tank during the day, or just make it an electric water heater (and then maybe expand the PV array a bit); it seems like electric/PV is a lot simpler and more flexible than solar thermal. Either way, the idea would be to control the water heater to run during the day and then drain the heat during the night, essentially acting as a storage tank. The boiler would still have to run in this scenario, but it can automatically lower combustion as low as 9K BTU. Raising a 100 gallon tank from 60F to 115F represents ~45K BTU of storage.
The radiant heat system right now is utterly dependent on electrical, so there would be the option to chuck my cockamamie ideas around (4) and just replace the Cadet LP boiler with a heavy duty electric tankless boiler.
(The water heater for the house is supplied by a separate LP tankless boiler, and it’s frustratingly in a completely different area of the house from the radiant heat mechanical closet.)
Thanks very much for any ideas and product recommendations. I’m fine spending a bit of money to get to much lower carbon footprint, but can’t embark on large-scale remodeling at this time.
GBA Detail Library
A collection of one thousand construction details organized by climate and house part
Replies
What are your water temps for the radiant?
What is your 99% outside design temperature? (Got a ZIP code?)
Reversible chillers with output ratings specified down to -13F exist, (eg Chiltrix, Arctic) and with low-temp radiant they can be 2.5-4x as efficient as an electric boiler at your seasonal average temperature.
https://www.chiltrix.com/documents/
https://www.arcticheatpumps.com/?gclid=EAIaIQobChMIu8aU9YaM4QIVRlYNCh3BsQZ_EAAYASAAEgK4wvD_BwE
Controlling the slab temperature with a floor thermostat and the room temperature with cold climate mini-splits can pretty much do the right thing for most of the year. By not raising the slab temp too high just to keep the house warm in the pre-dawn hours gives the slab's thermal mass more temperature range to work with when the solar gains start ramping, limiting the amount of temperature overshoot.
If it's been heated predominantly with the Lochinvar (and not so much the mini-split) you probably have enough data to estimate the 99% heat load using these methods:
https://www.greenbuildingadvisor.com/article/out-with-the-old-in-with-the-new
Code,
First of all, what is your name? (I'm Martin.)
Forget changes to your heating system. What you need is blower-door-directed air sealing, plus some cellulose added to the existing attic insulation (after the air sealing work is completed, of course).
If you do that, it's quite possible that your house will never freeze, even if you don't have a heating system.
Hi Martin & Dana,
I'm Jon. The house is about an hour west of Madison, Wisconsin, near Richland Center (53581). In extreme circumstances I think the heat load is a bit above 40K BTU, based on a 3–4 degree dip below the thermostat setting when we went to -25F this winter. But I think the load is normally in the 30–40K BTU range at night.
Dana, the Artic heat pumps with a buffer tank is kind of what I was imagining. Cooling with radiant would be bad, but maybe what makes sense in the long term is to see about replacing the current heat pump (primarily used for A/C in the summer) with a low-temp one that can run a fan coil unit for summer A/C and air-to-water for winter heating.
Martin, I hear you. I will get a blower door test scheduled for the next time I'm out there. The doors and primary windows don't see too leaky, but there are some secondary casement windows I have doubts about, as well as the vent stack in the attic. What do you think is a safe thermostat temperature? I've been conservative and kept it at 57 at the lowest, as some of the plumbing is in colder areas of the house.
Jon
Jon,
Like you, I tend to be conservative when it comes to thermostat settings, especially if you know that some of your plumbing is vulnerable.
Whether or not your house can "coast" through winter temperatures without heat depends on a lot of factors, as well as experience. New Passive House buildings routinely stay above freezing all winter long, without any heat. The tighter the home, the easier it is to protect the pipes. What you don't want are plumbing pipes near air leaks (for example, in a drafty basement or crawl space, near a faulty sill gasket that lets in outdoor air).
>" In extreme circumstances I think the heat load is a bit above 40K BTU, based on a 3–4 degree dip below the thermostat setting when we went to -25F this winter. But I think the load is normally in the 30–40K BTU range at night."
Your 99th percentile temperature bin in Madison is -6F, and the load at that temperature is the place to start design considerations, not some peak low number during the most recent Polar Vortex event or something. Run the fuel use based load numbers to get a handle on it. With a high gain house there will error on the peak load due to the offsetting solar gains, but it at least puts a firmer stake in the ground than " ...30–40K BTU range...".
A Manual-J type heat load calculation would probably be more accurate in your case than a fuel-use based number, but the fuel use based load number is still a relevant marker.
A Chiltrix or Arctic reversible chiller can run either heated or chilled water in fan coils (wall or floor mounted), but for best heating efficiency you'd want lower water temperatures than what's comfortable for a fan coil.
There's probably better bang/buck out of a floor wall mounted cold climate mini-split. A single 1.25 tonner like the Fujitsu - 15RLS3H is good for 16,000 BTU/hr @ -15F, and would cost about $4K all-in if put out to competitive bid in my area. Install it in the lower level, since you already have a heat pump for the upper level. Unlike the hydronic output chillers there is essentially no design work to getting the efficiency out of them. With chillers the water temperature going into the floor makes or breaks the heating efficiency. At your -6F outside design temp the COP efficiency of the Arctic heat pumps is about an output temp 95F, is a pretty-good 2.5, but that drops to under 2 at 115F, whereas the 15RLS3H is still over 2.5 even cranking along at it's maximum capacity (18,500 BTU/hr) at that outdoor temp, higher if modulating at output level less than that.
That's how controlling the slab temp with a chiller but controlling the room temp with a cold climate mini-split ends up being more efficient- lower water temps can be used if only keeping the slab a couple degrees above room temp. It also allows you to turn the slab heat off a few hours before dawn to manage solar gains. Cooling with just the mini-splits, not the chiller would be more efficient and comfortable too, with better management of latent cooling loads.
With a modulating heat pump like a mini-split working in tandem with the hydronic floor it's a "set and forget" sort of thing- the mini-split will modulate up/down based on the amount of solar gain is available, running at very high efficiency at part load. If opting for a chiller solution for the floor dedicating it to just the floor allows it to run at it's lowest possible water temperature for maximal efficiency. The thermal mass of the slab is usually/often sufficient to keep a modulating chiller like the Chiltrix from short-cycling, though dumber non-modulating chillers may still need a buffer tank.
But get a handle on the 99% load number using fuel-use methods first.
After a 44 day period in January–February, my LP tank was refilled with 209.5 gallons. That 209.5 gallons should be equal to 19.2 million BTU, an average of 436K BTU/day or 18.2K BTU/hr. That's not terrible, but obviously there's solar gain in that timeframe, and that 209.5 gallons of LP also resulted in about 1.3 tons of CO2 which is sobering.
So, regarding the 30–40K BTU heat load estimate, while it's not precise, I think it's accurate and I'll lay out my reasoning. The Lochinvar boiler is a condensing model with an efficiency rating of 94% and max output of 37K BTU. I haven't observed much/any short cycling. The Nest for the primary zone was installed throughout the entirety of January and reports having run for 446 hours (59% of the month) and 289 hours in February (43%—I was around for a week in Feb and used the wood stove, plus I got smarter about control and it was a somewhat warmer month). So, I know the boiler is mostly big enough to keep up with the heating load when needed and it maxes out at 37K BTU output, or I can guesstimate that the heat was on half the time; either way, the 30–40K BTU range seems fact-based and accurate, if not precise.
The current heat pump mini-split has a few problems. The biggest one is control, as it only has a remote control and has a minimum heating temperature of 63F. I need to figure out how to control it with a programmable thermostat. The other problem is that it's on the upper floor and not anything that really needs heat. It's a good place for A/C, but not for heat. Finally, I doubt it's a low temp model; I think it's a nothing-special Carrier, and probably at least 5 years old (there's some rust on the outside unit). Still, if I could solve the control problem, it could probably handle enough of the heat load to put a real dent in my LP usage.
I also don't have a problem dialing the radiant heat down to 95F. That's smart.
So, In the "short term" of what to do for next winter, it sounds like addressing heat loss (sealing + more attic insulation + shades for windows) and gaining control of the mini-split are the biggest things. Then I can measure throughout the winter and evaluate the mini-split/radiant electric/LP mix. In another ~5 years it might make sense to install a new heat pump system that's better able to handle both heating and cooling.
Thank you both for the detailed replies.
>"After a 44 day period in January–February, my LP tank was refilled with 209.5 gallons. That 209.5 gallons should be equal to 19.2 million BTU, an average of 436K BTU/day or 18.2K BTU/hr. "
The 18.2K is the AVERAGE load for the time period, not the peak load, not the load at the 99th percentile temperature bin for the location.
The average load for the period is useless for figuring out what the load is at -6F (your 99th percentile temperature bin) or -25F or any other temperature without also knowing the binned hourly outdoor temperature was for that period. (Or perhaps the outdoor temperature stayd rock steady at -6F all day and all night for that entire period? :-) )
The Nest data is similarly useless when dealing with modulating boilers, since the firing rate is unknown. The Nest can only track the duration and duty cycles of the calls for heat, and is completely agnostic of the state of the burner during those calls for heat. If it were a dumb on/off boiler with a single firing rate rather than a mod-con the duty cycle correlated to outdoor air temperatures could get you there.
If the Lochinvar is running under outdoor reset control, it throws yet another error factor into the mix, since the as-used combustion efficiency varies with both return water temperature and firing rate. Just because it tested at 94% in an AFUE test doesn't mean that's it's efficiency in your system, at your temperatures and firing rates.
Read the article, do the math on what the load is at your 99% outside design temperature:
https://www.greenbuildingadvisor.com/article/out-with-the-old-in-with-the-new
It's possible to make some corrections for use of auxiliary wood or mini-split heating for that time period based on fuel/electricity use, even if the mini-split isn't sub-metered, with some crunching of the power bills and the outdoor temperature for the period he mini-split was used.
Time frame: 1/15–2/28
209.7 gallons LP x 91600 conversion = 19,208,520 BTU
60F degree days: 1983.2
65F degree days: 2208.2
99% design temp: -6F
60F degree-day load: (19,208,520/1983.2)/24) * (60 - -6) = 26,635 BTU/hr
65F degree-day load: (19,208,520/2208.2)/24) * (65 - -6) = 25,733 BTU/hr
A couple of questions, though:
1) There's a week in this time frame where I was present, and the wood stove provided the bulk of the heat, thus the calculated heat load above is understated. So, should I throw out the degree days for that time period, or multiply the degree days for that time by a fudge factor (say, 15%, as the heat ran a little that week)? That probably puts me in the original estimated range of 30K–40K BTU.
2) Is this method applicable to my situation of wanting to provide for an average baseline temperature (in contrast with comfort)? I just care about the house not freezing when I'm not there, with a conservative safety margin. The wood stove provides for comfort when the house is occupied.
The boiler has a little flame icon on its display to show how much it's modulating. Although you're right that I don't have data, anecdotally I've never seen it at anything other than full throttle.
>"I’ve had this schedule in place for a few weeks now and it’s been better than simply having the heat set to 60; if it’s cold and not sunny, the Nest will start preheating the primary zone in the afternoon and run it until the 57F dropoff."
-----------
>"60F degree days: 1983.2
> 65F degree days: 2208.2
If you're keeping the indoor temps averaging 60F you should probably be using base 55F, then adding 5-10 heating degrees to get to what the load would be at a code minimum 68F indoor temperature capability.
You also used source fuel energy content of the propane, not the net output of the boiler that went into the heating system water. If the output water temps you're running are consistently under 120F you can use the 94% nameplate efficiency. If they're consistently over 140F use 87%. If it's in between call it 90%.
Do the math, but (doing the math in my head in my pre-caffeinated state) it'll come in around 35-37KBTU/hr using base 55F and projecting a few degrees for a 68F indoor temp.
And yes, you'll have to make corrections for the amount of wood used, and the efficiency of the wood stove. A load of 45-50K might even be real, but assume it's ~40K before you run an actual Manual-J on it. (And DO run a real load calc on it before buying any equipment.)
A single 15RLS3H would cover about half that @ -6F out/70F in. Assuming 95F water output a modulating Chiltrix reversible chiller running the radiant floor would be good for about 15,000 BTU/hr. ( https://www.chiltrix.com/chiller-technology.html ) Add another 15RLS3H and you'd be there with margin.
>"I just care about the house not freezing when I'm not there, with a conservative safety margin."
The Fujitsu mini-splits have a "MINIMUM HEAT" mode that puts it a a non-adjustable ~50F indoor setpoint precisely for applications such as this. At a 50F indoor temps they also have more capacity at low temps. You could just heat with the mini-splits, forget the radiant floor when you're not there, and use the Lochinvar to bring the slab up to some temp for comfort as-desired or as the "Hail Mary" backup for when it hit's -25F, saving the upfront cost of the Chiltrix.
edit: whoops, looks like Dana beat me to the punch!
"2) Is this method applicable to my situation of wanting to provide for an average baseline temperature (in contrast with comfort)? I just care about the house not freezing when I'm not there, with a conservative safety margin. The wood stove provides for comfort when the house is occupied."
Dana can clarify, but my understanding is that the HDD base temperature should equal your balance-point temperature. Your balance-point temperature is the outdoor temperature at which the heat loss through your building envelope equals the average heat gained from internal sources (appliances, bodies, etc.) and the sun. Your balance-point temperature will be lower if your thermostat setpoint is lower or your internal gains or solar gains are higher.
Dana's method is based on the assumption that your balance-point temperature is somewhere in the 60 F to 65 F range. That's a plausible range for a typical house with a the thermostat set for comfort (say, 68 F ish). If your thermostat was set lower during the period of your fuel use measurements, then you should use a lower HDD base temperature. If your setpoint was 58 F, say, then you might use 50 F and 55 F as the HDD base temperatures instead of 60 F and 65 F.
hi Jon -
I did not see this in the exchanges so far: setback strategies are notoriously difficult to coordinate/control with passive solar strategies. Radiant is slow while solar gain happens quickly. I agree with Dana; save radiant heat management for when you are there to be the "brain" behind the controls.
And my good friend Robert Bean would want me to add: don't reject radiant cooling because of a failure to manage indoor humidity. https://www.healthyheating.com/Biographies/Bean.htm
Peter
Thanks, Peter. The heat pump's FCU is on the upper level and unable to get much heat down to mechanical/plumbing systems on the lower levels; I also don't have good control over it at the moment. So, the radiant heat is my only option right now.
I've seen a few of the posts here about the trouble of matching radiant with passive solar. I agree! When I had the thermostats set to a particular temperature, say 60F, I would inevitably have the furnace running at sunrise, before the sun crested the hill, putting heat into the slab just when I wanted it at its low point. With new thermostats, I've set a schedule to back-off the temp (62 -> 57) at 2AM. I'm sure the schedule could use some fine-tuning, but it's been a great improvement as I don't have the heat running in the morning now.
The Nest's learning feature is worthless for my situation since it can't account for the solar gain. I would have preferred a cheaper wifi programmable thermostat, but the thermostats don't have C wires and the Nest can get along with trickle charging and I didn't have much time while I was there.
Jon, I'm in your area alot as an annual ticket holder to APT and the stories I've written about my work at Tower Hill State Park in your area's Voice of the River Valley. You can read my blog from 2016 here on GBA and look me up on the web. I'd love to help you look at some of the problems once you've done the blower door test.
Hey Paul! Awesome work on the brick oven at Tower Hill. I'm probably only a mile or two off your route from Appleton to APT. I'll be around the whole summer, so, yes, we should coordinate a convenient time for you to stop in before a show. At a minimum, I could use your help with figuring out the PV array, whether it makes sense to add more panels.
Since you know Devil's Lake, it wouldn't surprise me if you knew my house's previous owner and builder, Joel G, who died from cancer in 2013 and was on the Sauk County board.
Dana, Kevin — thanks for the clarification on the setpoint. I'll rerun the numbers. And Dana, thanks for the pointers to the Fujitsu and Chiltrix. The Chiltrix is interesting because they claim they can run any combination of indoor units off the same compressor. Having a couple of FCUs and a hydronic handler would be ideal. If I can cut the heat load with shades, more insulation, and sealing, then I might be within striking distance of a Chilltrix alone, especially paired with a buffer tank with a backup heating element.
I think it's unlikely that you can get the design heat load within striking distance of a Chiltrix, especially if using water temps north of 90-95F where it has less capacity. You might get it into the range of a 5 ton Arctic, but that would have a higher system cost and lower efficiency. With air coils and 90F water the wind chill of the tepid-air output makes it uncomfortable compared to the warm-air output of a mini-split operating it it's mid to high range. With higher temp water you run out of capacity quickly.
With high mass radiation like a radiant slab you won't need a buffer tank with the Chiltrix any more than you need one with the CDN040 boiler.
A Chiltrix only has a rationale if you're willing to pay a lot more up front for the luxury of a warm slab to cover a fraction of the load at reasonable efficiency. An electric boiler is cheap. 1.25 ton mini-splits are more efficient, and cheaper than building the system around any hydronic heat pump.
With a buffer tank and a big enough heating element you would be re-creating the electric water heater or high mass electric boiler (but at a higher price), which would use 2.5-3x as much electricity as a mini-split for the same amount of heat. Adding complexity to the radiant system also adds cost. In competitive bidding TWO Fujitsu 15RLS3H would run about $7500-7800 in my neighborhood (a single would run about $4K), and would deliver over 18,000 BTU/hr each @ -6F outside, 70F inside, 16000 BTU/hr each at -15F outside/ 70F inside (more at 50F inside). A pair of them puts out about as much heat at 70F indoor temps as the CDN040.
Putting one mini-split the lower level, another upstairs would pretty much cover it. The CDN040 could be replaced with a 10-12kwh electric boiler as the Hail Mary backup heat for about $1500 in hardware with about the same output capacity. Running the electric boiler off a floor thermostat and letting the mini-splits control the room temp when the woodstove goes out is a LOT more efficient than cranking up the slab heat with an electric boiler.
The one thing a buffer tank gets you is the ability to store some heat during the day and release it at night. That leverages the likelihood that electricity is cheaper during the day (from PV array) and the heat pump operating with more efficiency due to warmer outside air. It also tolerates mismatches in flow rate between the radiant system and the heat pump. Chiltrix's buffer tanks don't look too appealing, but Heat-Flo makes some storage tanks that look alright on paper (https://www.heat-flo.com/products/multi-energy-tanks).
To be sure, the Fujitsu looks good, too. I'm too ignorant right now to have an opinion on one heat pump versus another.
What I'd like to avoid, though, is having more than one condenser. I have an existing line coming into the house, with power, a place to put it, etc. Having two outdoor units would be a pain.
Also, not far away for great heat loss reduction on your windows: http://www.energywisemfg.com/
>"What I'd like to avoid, though, is having more than one condenser. "
You are saying you actively WANT a system that has a single point of failure?
The track record of cold climate Fujitsus operating on cold US climates are pretty good, but nothing is perfect.
There is almost no track record in the US on Chiltrix or Arctic hydronic heat pumps.
The biggest cold climate unit Fujitsu makes is a 3-tonner. Up to four xxRLS3 heads can go on a AOU36RLXFZH cold climate multi-zone compressor (it's only a condenser when in air conditioning mode), but the maximum output at -15F is 22,110 BTU/hr (which ~10KBTU/hr less than a pair of single zone 15RLS3H), and it's usually more expensive. It would be just one outdoor unit, but only ~6000 BTU/hr more capacity at -15F than a single 15RLS3H.
Mitsubishi has 3.5 and 4 ton cold climate multi-splits with more capacity. The 3.5 ton MXZ-5C42NAHZ is good for 33,800 BTU/hr @ -13F (slightly more than a pair of 15RLS3H). I'm not sure what the 4 tonner delivers, but I'm guessing it's around 35-36K @ -13F, and at your -6F
design temperatature would have comparable output to the CDN040. Like the Fujitsu the Mitsubishis have a mode where it can be set to 50F indoors for freeze protection. But the Mitsubishi units will shut themselves down at some temperature when the outdoor temperature sensor is out of range (which could be as warm as -18F), automatically turning back on when it warms up to -13F. That might not work for you during a Polar Vortex disturbance cold snap if you're not there to stoke the wood stove. But it does have the capacity.
Unlike cold climate mini-splits, Chiltrix doesn't have a pan heater and associated controls for automatically managing defrost ice build up over weeks of cold weather use. I don't believe Arctic units have those features either. (There isn't much online technical info on Arctic units.) That's fine if you're there to check the pan drain for ice plugs occasionally and clear it manually as-necessary, maybe even installing a heat tape to turn on for however long it takes to melt out any buildup. Lack of automatic defrost ice management not so great for remote operation, leaving it for a week or more at a time in the dead of winter.
A 10-12 kw electric boiler doesn't have that problem, but uses substantially more electricity.