Combined Heat and Power Systems: Cost, Spark Spread and Viability

A combined heat and power system is a capital project before it is an energy project. The engineering is well understood and the efficiency case is easy to make, which is why so many of these projects get as far as a quote and no further. What stops them is almost never the technology. It is the gap between two prices.

This page covers what the equipment costs, the arithmetic that decides whether spending it makes sense, and the gates a project has to clear on the way. If you want the mechanism first — what cogeneration is and how the heat gets recovered — start with cogeneration explained.

What a combined heat and power system consists of

The quoted machine is a minority of the project. A complete CHP system installation is:

• The prime mover package — engine or turbine, generator, enclosure, and the control system that synchronises it.
• Heat recovery — exhaust heat exchanger, and on an engine the jacket-water and oil circuits as well, plus the pumps and piping that get the heat to where it is used.
• Thermal interface — the tie into existing boilers, steam headers or hot-water loops, and the controls that decide which heat source serves the load at any moment.
• Electrical interconnection — switchgear, protective relaying, and the utility's own requirements for operating generation in parallel with its distribution system.
• Gas service — frequently an upgrade, because a continuous engine load is not the gas demand the building was sized for.
• Emissions and acoustic treatment — catalyst, silencers, stack, and whatever the air permit turns out to require.
• Civil and structural — pad, plant room or containerised housing, ventilation, and access for a machine that will need major overhauls on a fixed schedule.

Two consequences. The equipment line in a proposal is not the project cost. And the items most likely to move the total — gas service, electrical interconnection, acoustic treatment — are the ones that depend most on the particular building, which is why generic cost figures have to be read as a starting point rather than an estimate. Vendors describe the same arrangement as a CHP system, a CHP cogeneration package or simply cogeneration; the scope above is what each of those should include, and a quotation that omits several of these lines is quoting equipment rather than a project.

What it costs per kilowatt

The US Department of Energy publishes installed-cost figures by technology and size. Reciprocating engines, from the fact sheet DOE/EE-2764 of August 2023, on a net power basis:

Simple-cycle gas turbines, from DOE/EE-2841 of April 2024, on a gross power basis:

Three things to hold in mind when using these.

• They are 2020 dollars. Construction and equipment costs have not stood still since, and a 2026 quotation should be expected above the table rather than on it. Treat the figures as the shape of the curve, not the price.
• The two tables are not directly comparable. The engine figures are net of parasitic load and the turbine figures are gross, so a turbine's real cost per usable kilowatt is higher than the row suggests.
• DOE is explicit that site conditions dominate. Its own note is that installed costs are heavily influenced by site-specific requirements — which is the same point the scope-of-supply list above makes from the other direction.

What the curve does show unambiguously is that cost per kilowatt roughly halves across the size range. A 4.5 MW engine installation costs less than half per kilowatt of a 35 kW one. That single fact drives most of what follows.

Spark spread: the number that decides whether the project exists

Spark spread is the difference between what the utility charges you for a kilowatt-hour and what it costs you to make one. Everything else about a combined heat and power project is downstream of it.

The cost to generate has three parts: the fuel burned, minus a credit for the heat recovered, plus the incremental maintenance that running your own generation adds. The Department of Energy's screening material works an example in cents per kilowatt-hour — 5.45 for fuel, less 1.75 for the thermal credit, plus 1.90 for maintenance, giving 5.6 cents to generate. Against a utility rate of, say, 9 cents, the spread is 3.4 cents.

Whether that is enough depends on capital recovery. DOE's own illustration: a system at $2,500 per kilowatt, running 8,760 hours a year at 95 percent availability, carries about 3 cents per kilowatt-hour in the first year, falling to roughly 1 cent if recovered over three. Its stated rule of thumb is that the spark spread, net of operating and maintenance cost, needs to be on the order of 2 cents per kilowatt-hour to earn a reasonable return on the capital.

That threshold is the useful thing to carry away, because it reframes the whole exercise. A site with cheap electricity cannot be rescued by an efficient machine. A site with expensive electricity and no thermal load cannot be rescued either, because without the thermal credit the cost to generate rises and the spread closes. The machine is not the variable. The two prices and the thermal load are.

The screening questions worth asking before anything else

The Department of Energy's project-screening guidance reduces to three questions, and they are worth answering honestly before any money is spent on engineering:

• Do you pay more than 6 cents per kilowatt-hour on average, all in? That means generation, transmission and distribution together, not the commodity line.
• Does the facility operate more than 3,000 hours a year?
• Is there a thermal load present through the year, and does its shape match what the plant would produce? Both halves matter. A load that exists but peaks when the plant is idle does not help.

DOE is clear about what this screen is for: it decides whether a feasibility analysis is worth commissioning, not whether to build. A site that fails it does not need a study. A site that passes all three does, and that is a much better use of the first engineering dollar than a vendor's indicative quote.

Industrial cogeneration, and why scale changes the answer

Industrial sites do better at this than commercial ones for two reasons that compound.

They buy capacity more cheaply. The curve above halves across the engine range and falls by about 70 percent across the turbine range.

They convert fuel better. DOE's electrical efficiency figures rise with size — from 27.7 percent at 35 kW to 40.6 percent at 4.5 MW for engines, and from 22.3 percent at 2 MW to 34.7 percent at 25 MW for simple-cycle turbines. A larger machine makes more electricity from the same gas, which widens the spark spread from the cost side at the same time as the capital charge per kilowatt-hour falls.

Then there is duty. A manufacturing process that needs steam needs it whenever the line runs, which is the load profile these systems are built for. A commercial building's heat demand collapses for part of the year, and that is the shape the third screening question is asking about.

Worth correcting one common assumption, though: this is not only a large-plant technology. DOE reports that about three quarters of reciprocating engine systems in service are under 1 MW, and most are below 5 MW. A 500 kW installation at a site with the right load profile clears the screen comfortably. It is the load profile that matters, not the nameplate.

Where credit revenue fits, and where it does not

In Pennsylvania the metered output of a qualifying system earns alternative energy credits, sold separately from the electricity. It is worth being precise about how much that changes.

Pennsylvania's Tier II weighted average for reporting year 2024–25 was $26.92 per credit, which works out at about 2.7 cents per kilowatt-hour of generation. Set against a required spark spread on the order of 2 cents, that is not a rounding error — it is the same order of magnitude as the entire hurdle, and it can move a project that was sitting just the wrong side of the line. Compliance-year figures are on our Tier II price page.

But the mechanism matters, and it cuts one way only. Credits attach to generation, so they only accrue while the plant is running. A site that fails the thermal-load question does not run enough hours for the credit revenue to arrive, and no credit price creates operating hours. Credit revenue amplifies a viable project. It does not manufacture one, and a proposal that relies on it to reach the hurdle is a proposal to watch carefully.

The practical reading: screen on the spark spread first, and treat the credits as the thing that improves a return you had already decided was acceptable.

Two details that commonly surprise engineers:

• Pennsylvania sets no system-efficiency threshold for this purpose, unlike several other state programmes — the evidence is in Pennsylvania's AEPS program has no efficiency threshold.
• Credits run forward from the date a complete application is on file. A plant already operating has not been accruing anything, and the eligibility route and sequence are set out in how cogeneration projects qualify.

Where to go next

• Pennsylvania alternative energy credits for cogeneration — eligibility by system size, and a calculator that takes the operating hours this page argues you should question.
• Cogeneration explained — the mechanism, the prime movers and worked examples by sector.
• How cogeneration generates alternative energy credits — the credit calculation and the registration sequence.
• Measuring cogeneration output — what the meter counts, and why assumed run hours are the commonest modelling error.

Frequently asked questions