What Heat Recovery From a Compressor Can Realistically Deliver

Why this matters

Almost everything a compressor's motor draws comes back out as heat, and that fact gets sold as though the heat were free money sitting in a room. It is real, it is large, and shops that install recovery on the strength of the quantity alone are routinely disappointed, because the quantity was never the constraint. What limits recovery is the temperature the heat is available at and how much of it is rejected at moments when anybody wants it. A shop that can separate those three things - quantity, grade, coincidence - can tell a customer before anything is bought whether recovery on their machine returns a third of the heat or almost none of it, and can do it from a week of logging rather than a proposal.

Anything that taps a compressor's cooling circuit touches hot, pressurised fluid. Isolate the package at its electrical disconnect and its air outlet, lock and tag both under 29 CFR 1910.147, let the machine cool to where its lubricant cannot scald before you loosen anything, drain to a closed container, and confirm zero on a gauge open to the circuit before breaking a joint. Read the lubricant's safety data sheet first: compressor fluids include polyglycol and ester chemistries whose skin and eye controls differ from a mineral oil, and the sheet, not habit, selects the glove.

Where the energy actually goes

Track the electrical input to a package and there are only a few places for it to end up.

The great majority is rejected as heat: into the lubricant and through the oil cooler on an oil-injected machine, off the motor frame, off the airend casing, and through the aftercooler that drops the discharge air back toward room temperature. A small share leaves in the compressed air itself, because after the aftercooler the air is only a modest amount above ambient. A smaller share leaves with the condensate the aftercooler knocks out. There is no meaningful amount stored anywhere, because the air's pressure energy is spent in the plant, not in the compressor room.

So as a working statement: nearly all of the electrical input to a compressor package is available as heat somewhere in or around the package. The split between the oil cooler, the aftercooler and the casing varies by machine type and is on the package's own data, which is where you get it rather than from a rule of thumb.

For converting between the units you will be mixing: 1 kW is 3,412 BTU per hour.

Grade, not quantity, is the first constraint

Heat is only useful at a temperature above the thing you want to warm. The compressor's heat is available at the temperature of whatever medium is carrying it away, and that temperature is set by the machine's own design, not by how much heat there is.

On an oil-injected rotary screw, the lubricant leaves the airend hot enough to do useful work and is deliberately kept below the temperature at which the fluid degrades and the high-temperature protection trips. A liquid-to-liquid exchanger on that circuit can make hot water, but never hotter than the oil leaving the airend, less the exchanger's approach temperature. Ask the package manufacturer for the oil circuit's normal operating temperature and the exchanger vendor for the approach; those two numbers, subtracted, are your ceiling and everything downstream depends on them.

The consequence is a fork, and it is the fork that decides whether recovery is easy or marginal.

• Space heating wants air a modest amount above room temperature, which is far below any compressor's ceiling. The heat is comfortably above grade, so the whole question becomes coincidence.
• Process water above the ceiling cannot be delivered, only approached. That is not a failed project: preheating make-up water from cold to somewhere below the ceiling still displaces the same energy at the boiler. But the sales pitch has to say preheat, and a proposal that says the compressor will supply the process is wrong on its face.

The five things recovery cannot deliver

It cannot deliver more than the coincidence allows. The compressor rejects heat when air is being used. The building wants heat when it is cold. Those two schedules overlap partially and the overlap fraction, not the annual heat total, is the number that decides the project.

It cannot deliver a temperature above the machine's own operating temperature less the exchanger approach. No amount of exchanger surface gets you past it, and a bigger exchanger reduces the approach at a diminishing rate.

It cannot deliver a capacity credit. The existing heat source still has to be sized for the full load, because the compressor's output follows air demand and not heating demand, and on the coldest morning of the year the plant may not be running. Recovery is an energy saving, never a capacity saving, and a design that shrinks the boiler because recovery is present is the one failure in this list with a real consequence attached.

It cannot deliver across a mismatched architecture. An air-cooled package rejects its heat into the room and is recovered by ducting that air. A water-cooled package rejects its heat to a loop and is recovered at the loop. A recovery scheme designed for one and applied to the other returns nothing, and this is easy to get wrong on a site with both.

It cannot deliver a summer benefit into a conditioned space. Recovered heat dumped into a cooled building in July is a cooling load, and the mechanical cooling then pays to remove it. Every ducted-air scheme therefore needs a seasonal damper that diverts to outdoors, and that damper has to be checked, because one that fails in the recovery position in summer cooks the compressor room, raises the compressor's intake temperature and reduces the machine's own capacity, which the room ventilation and intake articles in this library cover.

The two architectures, and which one your package allows

Ducting the cooling air. The cheapest and the most common. The package's own fan pushes its discharge air into a duct that serves a warehouse bay or a make-up air path. The constraint is external static pressure: a package fan is rated for a limited amount of duct resistance, and exceeding it reduces the cooling airflow through the package, which raises its internal temperatures and can trip its high-temperature protection. Get the allowable external static from the package manufacturer before any duct is drawn, and where the duct run demands more, the answer is a booster fan interlocked to the compressor, not a longer duct on the package fan.

A liquid exchanger on the cooling circuit. More capital, more control, and it produces hot water that can be stored, which partly answers the coincidence problem by decoupling when the heat is made from when it is used. It also puts a heat exchanger into a circuit the machine depends on for its own protection, so it is manufacturer-approved work: a fouled exchanger on the oil circuit raises operating temperature and can shut the machine down, and a recovery loop that loses flow while the compressor runs does the same.

Worked example: the useful fraction, in hours and percentages

A plant runs one air-cooled rotary screw serving a two-shift operation, and wants recovered heat ducted into an adjacent warehouse bay that is currently heated by unit heaters.

Measure the input, do not calculate it. A logging power meter on the package for one representative week returns an average input of 26 kW across the hours the machine was energised, and the machine was energised 4,000 hours in the past year according to its own hour meter. Full-load input measured at the panel is 40 kW, which is worth recording because it sets the peak the duct has to carry: 40 kW times 3,412 is 136,480 BTU per hour at full load.

Annual heat rejected. 26 kW average across 4,000 hours is 104,000 kWh of heat put into that room over a year. That is the number the brochure quotes, and it is real.

Apply the availability fraction. The package data states the fraction of input energy that leaves in the cooling air stream the duct will capture, as against the casing losses and the small amount leaving with the compressed air and condensate. Say that figure is 75 percent, which is illustrative here and must come off your own package's data. 104,000 times 0.75 is 78,000 kWh available at the duct.

Apply the coincidence, which is the number nobody measures. The warehouse bay's unit heaters were logged for a heating season and called for heat during 1,700 of the compressor's 4,000 running hours. That is a coincidence fraction of 1,700 divided by 4,000, which is 42.5 percent. 78,000 times 0.425 is 33,150 kWh of recovered heat that actually displaces heat the bay would otherwise have bought.

State the result against the right base. 33,150 of 104,000 kWh is 31.9 percent, so just under a third of the heat this compressor rejects in a year ends up doing work. Not the "up to 90 percent" the quantity argument implies, and not zero. A third is a genuinely good result for a ducted scheme, and stating it as a third is what keeps the customer from feeling cheated in the second year.

Where the answer would move. Three conditions change that 31.9 percent materially. If the plant added a third shift, run hours would rise while heating hours would not, and the coincidence fraction would fall even as the absolute recovered energy rose slightly. If the bay were heated to a higher setpoint or were less well insulated, its call hours would rise and coincidence with it. And if a year-round heat sink existed, make-up water preheat being the usual one, the coincidence fraction stops being bounded by the heating season at all, which is why a liquid scheme on a plant with a constant hot water demand outperforms a ducted scheme on the same compressor.

How to know the coincidence number is real. Log it rather than assume it. A run-time logger on the unit heaters' control circuit and the compressor's hour meter, read over one full heating season, gives the 1,700 hours directly. Estimating it from a heating season's calendar length is the classic error, because it counts nights and weekends when the bay calls for heat and the compressor is off, and it overstates the fraction badly on a single-shift plant. Where the two schedules barely overlap, and a one-shift daytime plant beside a night-heated space is exactly that case, the honest recommendation is to spend the money on leak repair instead.

References

• 29 CFR 1910.147, control of hazardous energy, general industry: the basis for the isolate, lock, tag, cool and verified-zero sequence before opening a compressor's lubricant or cooling circuit.
• The compressor package manufacturer's technical data, for the split of input energy between the cooling air stream, the casing and the discharge air, for the oil circuit's operating temperature, and for the fan's allowable external static pressure. These are machine-specific and no rule of thumb substitutes for them.
• The lubricant's safety data sheet, for the skin and eye controls that differ between mineral, polyglycol and ester compressor fluids.
• See related: Why a Compressor Room Needs More Ventilation Than It Has; What Intake Conditions Do to Compressor Performance; Why Part Load Behaviour Decides the Energy Bill.