Winter and summer in Canada, you can do without paying for power.Year round, in hot climates or cold, you can skip the power bill completely.
Here is the build, what you need, and how it works โ plus all the legal information you should know, at the bottom.
And yes: this is a hobby project. Take the information and use it. It's free โ no warranty, no subscription, no company in the middle. Tag, you're it. If you build one, the best payment is publishing your numbers so the next person starts further ahead.
The design โ the intro
A few metres under any yard, the ground holds the year's average temperature, steady, all year โ about +5 ยฐC around Edmonton โ while the surface swings from โ30 to +30. That gap between the surface and the depths is free energy, and it is biggest exactly when energy costs the most.
Drill a field of narrow holes, about 15โ20 m deep. Put a sealed loop of pipe in each one and grout it in. Lay one plate over the whole field and let the grass grow over it. Then:
A heat pump moves the earth's steady warmth into the house โ 3 to 5 units of heat for every 1 unit of electricity it spends. This is the big win, and it's how the power bill dies.
The house's fresh air gets pulled across the field first, so it arrives near ground temperature instead of โ30. In deep winter that slice alone is worth 15โ30% of the heating load. No machine involved.
A small recovery unit โ one sealed CO2 loop and a little expander machine โ turns the same gap into a trickle of electricity, day and night, weather or no weather. Not a lot. Enough for the things that matter, through any outage, forever.
In summer the system banks heat back into the ground while it cools the house, recharging what winter withdrew โ so the field doesn't run down.
Five pieces, from the ground up:
The field. Holes ~150 mm (6") across, 15โ20 m deep, spaced ~5โ6 m apart. The 6 inches is the hole โ the ground, not the pipe, is the bottleneck, and a wider hole buys almost nothing while drilling cost climbs with the square of the diameter. Each hole holds one U-loop of 32 mm pipe. Thirty holes fit in an ordinary backyard corner โ about 510 m of borehole, a whole house with margin.
The fill. Thermally enhanced grout โ bentonite loaded with silica sand โ pumped in from the bottom up in one continuous pour.
The plate. One large interconnecting aluminium plate over the field: shared manifold, armour against future shovels, and a heat exchanger in its own right. Soil and grass over the top. The field takes zero land and disappears.
The machines โ all topside. Nothing expensive is ever buried. Circulation pump, reversing valve, sensors, and the recovery unit live in an inspectable valve box at the plate. The heat pump is a standard unit indoors.
Storage. Capacitors in front (they take every trickle watt instantly, almost no loss), lithium-iron-phosphate batteries behind them for bulk. Surplus in winter gets dumped as house heat โ nothing wasted.
The materials โ what you need
Almost everything on this list already exists on a shelf somewhere. Only one part has to be built.
Loop pipe: 32 mm HDPE, on coils. Any plumbing/geo supplier. Two legs per hole (a double-U of four legs fits too).
Grout: bentonite + silica sand, mixed as thermally enhanced grout (~1.4โ2.0 W/mยทK, about double plain bentonite). Local bulk โ too heavy to ship, sold everywhere.
Drilling: a 6" sectional bit and a trailer-mounted press-and-spin rig (rent, build, or hire the crew). The bit tip is the wear part; it gets replaced as it wears.
The plate: aluminium sheet/section, with isolators anywhere it meets a different metal. Aluminium doesn't rust and conducts heat 3โ4ร better than steel โ the plate is a heat exchanger, not just a lid.
Fittings and valves: standard refrigeration/geo shelf โ unions at every joint you'll ever want to open, isolation valves, and a heat-pump reversing valve for the season flip.
Circulation pump: a small standard circulator.
Heat pump: an ordinary ground-source (geothermal) unit from the existing industry. This design doesn't reinvent it.
Sensors: temperature at the top and bottom of the field, pressure on both sides of the recovery unit. Cheap, standard.
The recovery unit (the one built part): a sealed steel CO2 loop and a small positive-displacement expander-generator, built as a lift-out cartridge.
First test article: a hermetic fridge compressor โ nearly free, pressure-rated, no shaft seal to leak.
The serious version: an inverter scroll AC compressor run backwards. Factory-built for variable speed.
The supply shelf: European supermarkets run CO2 refrigeration at these exact pressures (90โ120 bar class), so rated valves, driers, fittings, and compressors are catalogue parts (Bitzer/Dorin class).
Safety parts, not optional: pressure-relief valve + burst disc sized for a sealed charge baking in summer sun.
Controller: an MPPT-style DC-DC controller โ the solar trick โ feeding the storage bus.
CO2: food/welding-grade, from any gas supplier. Bought locally; pressurized gas doesn't ship in kits.
Storage (yours to choose, spec provided): a capacitor bank + LFP batteries + an inverter sized to your house. The kit suggests; you decide.
Ducting: sheet metal for the intake-air path across the plate.
How it fits together, then how to build it
The ground is the battery and the loops are the wires. Each loop is a straw into the steady temperature: fluid goes down one leg at surface-influenced temperature and comes up the other at ground temperature. The plate gathers all the straws into one manifold. The heat pump drinks from the manifold. The recovery unit sits across the CO2 loop where the temperature โ and therefore pressure โ difference is, and skims electricity off it. The intake duct crosses the plate so incoming air trades its temperature with the surface before it ever reaches the house.
Stage 0 โ One hole, one winter (do this first, always)
Drill ONE hole. One loop, grouted, sensors top and bottom, and log it for a season. This costs little and tells you the truth about your ground, your grout pour, and your drilling before you've committed to thirty holes. Publish the log. Every number in this design gets tested here first.
Stage 1 โ The field and the heat
Locates first. Always. Utility locates before any drilling (in Alberta: Alberta One Call โ free), and respect setbacks from lines and property lines.
Drill the field. Drop each loop, hold it down so it can't float in the grout, and grout bottom-up through a tremie pipe, one continuous pour per hole. Air voids are the real enemy โ a careful pour matters more than any other step.
Lay the plate/manifold, connect loops with unions, pressure-test every loop. Expect roughly 9 mยณ of spoil for a 30-hole field; plan where it goes first.
Have a ticketed installer connect the heat pump and a licensed electrician do the panel work. (See Legal, bottom.)
Stage 2 โ Intake air
Duct the ventilation intake across the plate, with a damper arrangement for the spring/fall crossover weeks when there's little gap to harvest. Sheet metal and patience; worth 15โ30% of the ventilation slice of a winter bill.
Stage 3 โ The recovery unit (the experiment)
A sealed two-phase CO2 loop: CO2 evaporates at the warm end, condenses at the cold end, and the pressure difference drives the little expander. CO2 is the right fluid for this gap โ its critical point (31.1 ยฐC) sits right in the working range โ it's cheap, and it's sold everywhere. Safer first prototype: build the sealed steel assembly and pressure-test it on the bench, then lower it in.
Its spec, in the designer's words: it has to take everything variably โ take every bit of pressure when it can, and not waste.
Positive displacement, not blades. A small turbine is only efficient at the one condition it was designed for. This machine faces everything from zero (crossover weeks) to the full winter gap, reversing seasonally. Scroll, piston, and vane machines hold efficiency across wide pressure ratios, start on a trickle, and swallow the liquid slugs CO2 throws near its critical point.
Waste nothing. At this gap, every loss is proportionally fatal โ leakage, fitting pressure drops, friction. Big ports, short paths, tight clearances, and no throttling valve anywhere: the pressure drop happens across the machine, doing work, or it doesn't get spent at all.
Absorb variability twice. Mechanically, variable cutoff โ steam-engine tech, a century old, invented for exactly this. Electrically, let it spin at whatever speed the pressure wants and take the power at its best point with the MPPT controller, into the capacitor/battery bus.
Pressures are high, ratios are gentle. About 17 bar at โ25 ยฐC, 40 bar at +5 ยฐC, up to ~72 bar at +30 ยฐC โ a ratio of only ~2:1. The worst case is a sealed charge on a sun-baked plate; the fill amount is the safety knob, and the relief valve and burst disc are mandatory.
Keep it dry inside. Any moisture in a CO2 loop becomes carbonic acid and eats the machine from within. Deep vacuum, a drier, sealed assembly. Boring and absolute.
Repair means replace
The recovery unit is one sealed cartridge on unions, isolation valves both sides, a bypass loop so the heat side runs without it. Failure = close valves, unbolt, swap, minutes. The dead one gets bench-repaired indoors or not at all. Keep one spare cartridge and an envelope of CO2-rated seals on the shelf. Charge work is standard refrigeration practice: recover into a bottle, open, recharge by weight.
What it actually does โ the honest part
Heat: pays for the project. 3โ5 units of heat per unit of electricity, from ~150โ250 m of borehole for a typical house. This is proven technology โ a mature industry does it every day.
Electricity: a trickle with perfect timing. Roughly 10โ30 W per hole in a strong gap โ a 10-hole field is 100โ300 W, 24/7, weather-independent. Lights, fridge, freezer, phones, internet: the essentials tier, including through a January grid outage. It does not replace the grid for a whole modern house on the electric leg alone. January watts are the expensive watts, and this makes them when nothing else does.
Intake air: free money in the extremes, nothing in the crossover. The whole design feeds on the gap; it earns at the extremes of the year.
No fluid and no cleverness beats the physics of a small temperature gap. If someone promises you kilowatts of electricity from a backyard temperature difference, they're selling something.
How many holes run a house โ and a car
One hole is a trickle. A field is a different animal โ the watts add, hole by hole, 24 hours a day. The question is how many holes, and the honest answer has two ends, because a hole makes 10 W at the low end of a strong gap and 30 W at the high end. These are strong-gap numbers โ deep winter, when the gap is biggest. In the shoulder seasons the gap shrinks, and so does the electric leg.
The budget, as typical estimates: a house's non-heating electricity โ lights, fridge, freezer, phones, internet, laundry, the rest โ runs roughly 10โ15 kWh a day. A car driven daily and charged at home adds roughly 10 kWh a day. Together, about 20โ25 kWh a day the field has to supply, averaged over 24 hours. Heating is deliberately not in that budget: the heat pump carries heating at 3โ5 units of heat per unit of electricity, and that is the whole reason house-plus-car is in reach at all. Trying to heat electrically too would add tens of kWh a day in an Edmonton winter, and no backyard field covers that on the electric leg.
10 holes โ 100โ300 W, 2.4โ7.2 kWh a day. The essentials tier: lights, fridge, freezer, phones, internet, through a January outage.
30 holes โ 300โ900 W, 7.2โ21.6 kWh a day. An ordinary backyard corner, about 510 m of borehole. At the high end, the house's non-heating load plus a daily car charge. At the low end, roughly a third of that budget โ essentials plus a good share of the rest.
45 holes โ 450โ1,350 W, 10.8โ32.4 kWh a day. House and car at the high end with margin; about half the budget at the low end.
So, plainly: at the high end of the range, a field in the 30โ45 hole range powers the house and the car in a strong winter gap. At the low end, the same field doesn't โ full house-plus-car would take on the order of 100 holes, more land than a normal yard has. That spread is exactly why Stage 0 comes first: drill one hole, log a season, and your ground tells you which end of the range you're on before you commit to thirty holes.
Storage: capacitors first, good batteries behind โ and why cheap costs more
The engine's output is variable โ pressure follows the gap hour to hour โ and the turbine must never be throttled, because a valve drop is pure waste. So storage is part of the machine, not an accessory bolted on after.
Capacitors first. They take every watt instantly at almost no loss, smooth the variation, and let the controller take power at the turbine's best point. They also keep thousands of little charge/discharge micro-cycles off the batteries โ micro-cycling is what wears a battery out early. Supercapacitors last on the order of 100,000 cycles, but they self-discharge over days: they are the buffer, not the store.
Good batteries behind. Lithium-iron-phosphate: roughly 90โ95% round-trip efficiency, thousands of cycles, and about 80โ90% of the nameplate actually usable. They hold the field's output for days, which is what a car charge and a slow week need.
The waste issue, in numbers. Losses multiply down the chain. A good chain โ capacitor ~98%, LFP ~95%, good inverter ~95% โ delivers about 88% of the harvested watt to the load. A cheap chain โ capacitor ~98%, cheap lead-acid ~75%, cheap inverter ~85% โ delivers about 62%. On a 20 kWh harvest day, the cheap chain throws away roughly 5 kWh more than the good one: a quarter of the field's output, every day. And the cheap bank needs replacing every couple of years โ a few hundred cycles, and only about half its nameplate usable, so you buy twice the bank for the same store.
Before going for the cheapest option, price storage per delivered kWh over its life, not per nameplate kWh on the day you buy it. The cheapest battery is usually the most expensive part of the system โ and it wastes the watts the field worked all winter to make.
The kit (if you productize it)
The designed core only โ everything else is deliberately somebody else's shelf, which is why the kit can cost next to nothing:
- In the box: loop sets, the plate/manifold with valve box, the recovery cartridge (shipped uncharged), sensors + controller, and the paper โ hole pattern, spacing, fill spec, and a suggested spec for the buyer's inverter/batteries/capacitors.
- Not in the box: drilling (local labour โ a crew can make a living at it), grout and CO2 (local bulk), the heat pump (the existing industry's), storage electricals (the buyer's).
- Sell it just above cost if you sell it. The shape that fits: the designer sells the core; diggers dig; a fill route develops the way propane routes did; electricians wire. Nobody owes anybody a service contract.
What's already proven, and what isn't
Proven elsewhere: borehole-field heat pumps (mature industry); earth-tempered intake air (decades of use, standard in Passivhaus design); CO2 thermosiphons (the Alaska pipeline rests on them); seasonal borehole storage like the banking here (Drake Landing, Okotoks, Alberta โ 52 homes, 144 boreholes, solar-charged, ran ~17 years and published its numbers).
Not done anywhere, as far as we can find: an engine leg running on a ~25โ30 K gap, and this exact assembly โ field, armour plate under sod, intake, pump, CO2 engine, capacitor/battery bus โ as one design. The pieces are proven. The combination and the engine are open ground. The tests are cheap, the truth is knowable in one winter, and the heat half pays you back while you find out.
Legal โ the information you should know
This is a hobby build while it's your land (or land you have the owner's full, informed consent for), your labour, no customers, no sale of power, no grid export. Those are the lines โ not the number of holes.
Locates are not optional. Utility locates before drilling, every time. Free, fast, and the law in most places.
Groundwater: closed loops draw no water and touch no wells โ but boreholes reaching groundwater can pull licensed-driller requirements depending on your jurisdiction. One phone call to your provincial/state regulator before Stage 1 settles it.
Permits: the heat pump connection is a mechanical install โ permit it like a furnace swap. Electrical work by a licensed electrician. Unpermitted heating or electrical work is an insurance problem even when it's a safety non-issue; your insurer is the objector with real teeth. Tell them.
Pressure equipment: the CO2 version contains a pressure vessel, and pressure vessels are regulated (in Alberta: ABSA territory) the moment it's a product or in someone else's yard. For your own bench prototype, respect it as the stored energy it is: relief valve, burst disc, pressure test before it ever goes in the ground. The plain water/glycol loop version has no pressure boundary at all โ one reason it's the recommended first build.
The CO2 top end is the corrosion weak point. Buried, grouted steel is protected (alkaline grout passivates steel; the inside of a sealed, dry CO2 tube is oxygen-free and self-protecting). The air-exposed top hardware is where rust would live โ so bury the cap below grade, and keep reachable fittings brass or stainless, small, inspectable, swappable. Worst case after decades of neglect: a slow leak of a benign gas and a dead system. Not a hazard event.
The landowner is the only true veto. If it isn't your yard, their informed yes is the whole foundation. Either way, leave a note with the house records: there is a field here โ don't dig up the flower bed.
It stops being a hobby build if you sell power, install for others, or export to the grid (export needs an interconnection agreement with the utility; your own loads never involve them). Installing for others makes you a contractor โ tickets, insurance, warranties, callbacks. Price that honestly, or publish the design and let ticketed trades build it. This document is the second door.
Designed by Apu (Aly V.) with Muse, October 23, 2026, Edmonton, Alberta. Released freely the same day. Build it, improve it, beat it โ and publish your temperature log. The numbers are the point.
Buy me a coffee
This project is free, and it always will be. If it helped you, a coffee is appreciated — it helps keep the builds going.
Thank you! — skooby99@outlook.com
๐ฌ Feedback
Found a bug, or have an idea to make this better? Leave feedback here — report issues and vote on what should be fixed next.