Coal, Gas and Combustion: The Incumbent and Its Economics

10.1 — Combustion is a chemistry problem in engineering clothing

Burning a fossil fuel is an oxidation reaction. Carbon and hydrogen locked in the fuel combine with oxygen from the air, and the products — carbon dioxide and water — sit at a lower energy state than the reactants. The difference is released as heat.

Everything else in a thermal power station is machinery for turning that heat into rotation. And because it is heat, Chapter 3’s constraint applies without appeal: a heat engine cannot convert all of it. The Carnot limit — one minus the ratio of the cold absolute temperature to the hot one — is not an engineering target that better materials will eventually beat. It is the ceiling that every combustion plant is built underneath.

Which leaves engineers exactly two levers: raise the hot temperature, or lower the cold one. The cold end is the ambient environment — a river, the sea, the air — and is essentially fixed near 30–40 °C. So the entire history of thermal power plant development is the history of firing hotter without the machine melting.

10.2 — The Rankine cycle: coal’s heat engine

A coal plant does not send combustion gases through the turbine. It burns coal in a furnace to heat water in tubes lining the walls, converts that water to steam, superheats the steam well above its boiling point, expands it through a turbine, and then condenses it back to water so a pump can return it to the boiler. Water is the working fluid and it never leaves the loop; the fire is on the outside.

This is the Rankine cycle, and the reason for the architecture is practical rather than theoretical. Coal combustion produces ash, sulphur, chlorine and abrasive particulates. Nothing you would willingly pass through a precision turbine. Separating the fire from the working fluid means the turbine sees only clean steam, at the cost of an extra heat-transfer step and everything that comes with it: a boiler the size of a building, hours to warm up, and a hard ceiling on temperature set by what the boiler tubes can survive.

The condenser also explains the plumes. What leaves a cooling tower is water vapour carrying away the heat the cycle could not convert — roughly two units rejected for every unit of electricity produced. The chimney is a separate structure carrying combustion products, and it is much less visible.

10.3 — Subcritical, supercritical, ultra-supercritical

If the only lever is temperature, the whole engineering programme becomes: push steam conditions higher. Above 22.1 MPa and 374 °C water reaches its critical point, beyond which liquid and vapour stop being distinguishable — there is no boiling, no bubble, no phase boundary, just a single dense fluid. A plant operating above that point is called supercritical, and the terminology of the industry follows directly.

The pattern in Figure 10.1 is worth sitting with. Moving from subcritical to ultra-supercritical raises the Carnot ceiling by only about three percentage points — from 61.5% to 64.9% — because absolute temperature is measured from −273 °C and 540 °C to 620 °C is a modest fractional change. Nearly all the real gain comes from capturing a larger share of that ceiling: 59% of Carnot for the subcritical unit against 69% for the ultra-supercritical one, won through better blade aerodynamics, reheat stages, feedwater heating and lower auxiliary loads.

10.4 — The Brayton cycle: the gas turbine

A gas turbine takes the opposite approach. Air is drawn in, compressed to fifteen or thirty times atmospheric pressure, mixed with fuel and burned continuously, and the hot high-pressure gas expands through turbine blades on its way out. There is no boiler, no steam and no condenser. The combustion products are the working fluid.

That is the Brayton cycle, and it is the same machine that powers a jet aircraft — many industrial gas turbines are directly derived from aero engines. Two consequences follow immediately, and they define the technology.

• •It can fire far hotter. Turbine inlet temperatures of 1,400–1,600 °C are routine, against 540–620 °C for steam, because the blades are cooled internally by compressor air and coated with ceramic thermal barriers. The gas is hotter than the metal’s melting point; the metal survives because it is being actively refrigerated.
• •It needs a clean fuel. Combustion products pass straight through the turbine, so ash and sulphur are not merely undesirable, they are disqualifying. This is why gas turbines burn natural gas or distillate and coal does not go anywhere near one without first being gasified.

Yet an open-cycle gas turbine reaches only about 38% — barely better than a subcritical coal plant, despite firing nearly a thousand degrees hotter. Figure 10.1 shows why. The exhaust leaves at roughly 550 °C and goes up the stack. The machine has an excellent hot end and a catastrophically bad cold end, and Carnot punishes the cold end just as hard.

10.5 — Combined cycle: two engines in series

A combined-cycle gas turbine routes the exhaust into a heat recovery steam generator, which raises steam, which drives a second turbine, which rejects to a condenser at ambient temperature. The Brayton cycle sits on top; the Rankine cycle sits underneath and picks up what the first one discarded. Roughly two-thirds of the output comes from the gas turbine and one-third from the steam turbine.

The result is the most efficient prime mover ever built for grid electricity: about 62% net for a modern unit, with the best machines quoted above 64%. Figure 10.1 shows the mechanism precisely — the hot end stays at gas-turbine temperature while the cold end drops from a 550 °C stack to a 40 °C condenser, and the Carnot ceiling jumps from 51% to 83%.

The same logic extends further. In combined heat and power, the steam cycle’s low-grade reject heat — useless for making electricity — is sold as district heating or process steam, pushing total fuel utilisation past 80%. That is a different measure from electrical efficiency and the two should never be compared directly, but the principle is the same: stop rejecting heat that someone downstream can still use.

10.6 — Heat rate: the industry’s currency for efficiency

Thermal efficiency is a physicist’s number. The number traded, contracted and dispatched on is its reciprocal, called heat rate: the fuel energy required per unit of electricity out.

A 36% subcritical coal plant has a heat rate of 10.0 GJ/MWh. A 45% ultra-supercritical unit needs 8.0. A 62% CCGT needs 5.81. In markets that quote imperial units the same numbers appear as 9,478, 7,583 and 5,506 Btu/kWh. Lower is better, which catches people out the first time.

Two cautions. First, quoted heat rates are usually for new and clean equipment at full load; real plants degrade with fouling, wear and seal leakage, and a unit held at 50% load can be several percentage points worse than its nameplate. Second, fuel energy can be quoted on a higher or lower heating value basis — the difference is whether the energy in the water vapour of combustion is counted — and for natural gas the two differ by roughly 10%. A 60% LHV efficiency is about 54% HHV. Comparisons across jurisdictions frequently mix the two and produce arguments about nothing.

10.7 — Fuel cost per megawatt-hour

With a heat rate and a fuel price, the number that governs everything else falls out immediately.

10.8 — The merit order and who sets the price

A wholesale power market dispatches plants in ascending order of short-run marginal cost until supply meets demand. That ordering is the merit order, introduced in Chapter 4 and now filled in with the actual numbers from the previous section. Zero-fuel renewables sit at the bottom, then nuclear, then coal, then gas, then peaking plant.

The rule that makes this matter is marginal pricing: every dispatched plant is paid the offer of the last plant required, not its own. So the market price is set by the most expensive machine running, which in most systems most of the time is a combustion plant — even where combustion supplies a minority of the energy.

The gap between a plant’s own marginal cost and the clearing price is infra-marginal rent, and it is not a windfall in any pejorative sense — it is how capital-intensive plant recovers capital in an energy-only market. It is also volatile: in Figure 10.3 the same fleet earns about $0.5m an hour of rent at off-peak demand and $3.3m at peak, with no change in any plant’s cost. A solar farm’s revenue therefore depends on a gas price, and its timing depends on when the expensive plant is needed — which is exactly the mechanism behind the value deflation discussed in Chapters 6 and 7.

10.9 — The screening curve

Marginal cost decides which plant runs tonight. A different calculation decides which plant gets built, and it is the reason generation fleets contain several technologies rather than only the cheapest one.

Take the annual cost of owning one kilowatt of capacity — capital repayment plus fixed operations and maintenance — and add the cost of running it, which depends on how many hours it runs. That produces a straight line per technology: intercept set by capital, slope set by fuel. Plot all of them together and you have a screening curve.

The screening curve also explains why every crossing point is a hostage to assumptions. Raise the cost of capital and the capital-heavy technologies move up; raise the gas price and the gas lines steepen; add a carbon price and the coal line steepens faster than either. This is the same LCOE machinery as Chapter 4, but plotted so the sensitivity is visible rather than buried in a single quoted number.

10.10 — Three roles: baseload, mid-merit and peaking

These are economic categories, not engineering ones. Nothing prevents a CCGT running 5% of the year; it is simply an expensive way to own a peaker. And the categories are not stable — as renewables displace energy without displacing capacity, plants migrate down the table. A coal unit built as baseload and now running 40% of the year has been recategorised by the market as mid-merit while its balance sheet still assumes baseload.

10.11 — Minimum load, ramping and the cost of turning down

A thermal plant cannot be turned down to zero and left running. Below a minimum stable load the flame becomes unstable, steam temperatures fall out of specification and emissions control equipment stops working properly. Typical floors are 40–50% of rated output for older coal, 25–35% for modern coal, and 20–40% for a CCGT depending on configuration.

Two consequences follow, both increasingly important as renewable penetration rises.

• •Efficiency falls at part load. A plant at minimum load may be three to six percentage points less efficient than at full load, so its heat rate — and therefore its marginal cost and its emissions per megawatt-hour — get worse exactly when it is producing least.
• •Minimum load competes with renewables. A coal unit that must stay at 40% to be available for the evening peak is occupying 40% of its capacity in the middle of a sunny afternoon. In systems with a lot of must-run thermal plant, this is a direct cause of solar and wind curtailment — the renewable energy is free and it is thrown away because the thermal fleet cannot get out of the way.

Ramp rate is the companion constraint: how fast output can change once running, usually quoted in percent of capacity per minute. Coal manages roughly 1–3% per minute, a CCGT 4–8%, an open-cycle machine 10–20%, and a reciprocating engine or battery effectively instantaneously. As net load ramps steepen — the duck curve of Chapter 5 — ramp capability becomes a separately valuable product, and a slow plant can be technically available and commercially useless.

10.12 — Start-up cost and cycling damage

Start-up costs are not mainly fuel. They are the accounting for damage. Every thermal cycle puts a plant’s thick-walled components through expansion and contraction, and repeated cycling consumes creep-fatigue life in headers, drums, rotors and casings. A unit designed for a few hundred starts over thirty years and now performing three hundred a year is spending design life at roughly thirty times the intended rate.

10.13 — Carbon dioxide from first principles

Carbon intensity does not have to be looked up. It follows from the chemistry of the fuel and the efficiency of the plant, and deriving it once makes every quoted figure in the field readable.

Hydrogen content is the other half of the story. Coal is mostly carbon; natural gas is mostly CH₄, and roughly a quarter of its combustion energy comes from oxidising hydrogen to water rather than carbon to CO₂. Gas is not a cleaner fuel because it is handled better. It is a cleaner fuel because a meaningful fraction of what is burning is not carbon at all.

10.14 — The other emissions

Carbon dioxide dominates the climate discussion and almost none of the public health one. The pollutants that kill people are local, and the control technology for each is mature, expensive and specific.

Three points are worth carrying forward. Control equipment consumes energy, so a fully controlled plant has a worse heat rate and higher CO₂ per megawatt-hour than an uncontrolled one — cleaning the local air costs a little climate. Gas turbines produce essentially no SO₂, particulates or mercury, but they do produce NOₓ, because that comes from the air rather than the fuel. And the health burden is overwhelmingly a coal burden: on any accounting of deaths per unit of energy, coal is orders of magnitude worse than gas, which is in turn worse than anything in Chapters 6 to 9.

10.15 — Water, and the constraint nobody prices

A steam cycle needs a cold end, and the cold end needs a heat sink. For a thermal plant that means water, air, or an expensive compromise between them.

10.16 — Upstream methane

Everything in Section 10.13 measured emissions at the stack. For gas, that understates the total, because methane leaks between the wellhead and the burner tip — at wells, gathering lines, compressors, processing plants, pipelines and distribution networks.

Methane is a far more potent greenhouse gas than CO₂ while it lasts: roughly 30 times over a hundred-year horizon and 80 times over twenty years, because it oxidises in the atmosphere within about a decade. Which horizon is used changes the conclusion, and the choice is a value judgement rather than a scientific one.

10.17 — Carbon capture and the energy penalty

If combustion is going to continue, the obvious move is to catch the CO₂ before it leaves the stack. The chemistry works and has for decades: flue gas is contacted with an amine solvent that absorbs CO₂, the solvent is heated to release it, and the concentrated gas stream is compressed and injected underground.

The difficulty is that both steps — regenerating the solvent and compressing the gas — consume energy, and that energy comes out of the plant’s own output. The energy penalty is typically 8–12 percentage points of efficiency for post-combustion capture on a CCGT, and more on coal because there is more CO₂ per megawatt-hour and it is more dilute in the flue gas.

The economics compound the same way. Capture adds capital, adds fixed O&M, and raises the fuel bill per unit sold, so it makes an already fuel-heavy technology more fuel-heavy — and then requires a CO₂ transport and storage system with its own capital, its own liability and its own permitting. Capture is most defensible where the CO₂ stream is already concentrated and there is no alternative process route, which describes cement and some industrial applications far better than it describes power generation competing against $30/MWh solar.

10.18 — The carbon price that switches coal to gas

Fuel switching is the fastest emissions lever available to an existing power system, because it requires no construction at all — only a change in dispatch order. And the price at which it happens is a four-line calculation.

10.19 — Why plants stay open

Given the arithmetic above, a reasonable person would expect uneconomic coal plants to close promptly. They frequently do not, and the reasons are worth understanding because they recur across every asset class in this guide.

• •Sunk capital is sunk. A plant whose construction cost is already spent only needs to cover its going-forward costs to be worth running. It can be a disastrous investment and a rational operation simultaneously.
• •Contracts outlive economics. Long-term power purchase agreements and take-or-pay fuel and rail contracts can oblige payment whether or not the plant runs, which makes running it the cheaper of two bad options.
• •Capacity payments. Where a market pays for availability rather than energy, a plant can earn most of its revenue from being ready and very little from generating.
• •Reliability obligations. A system operator may simply refuse to permit closure if the unit is needed for local voltage support, transmission constraints or winter adequacy — sometimes with an out-of-market payment attached.
• •The rest of the balance sheet. Closure crystallises decommissioning, site remediation, ash pond liabilities and redundancy costs immediately, while continuing to operate defers them.
• •Employment and regional politics. Coal plants and the mines feeding them are often the economic base of a specific region, and that is a real consideration, not an irrational one.

10.20 — Gas as the bridge, and the residual load argument

Gas is routinely described as a transition fuel. The claim has a strong version and a weak version, and they are worth separating.

The strong version is the substitution argument: replacing coal generation with gas generation roughly halves emissions per megawatt-hour immediately, using proven technology and existing supply chains. Section 10.16 is the caveat — the halving is closer to a 2.3× reduction once upstream methane is counted, and worse on a twenty-year warming horizon — but the direction is not in dispute.

The weak version, and the more durable one, is the flexibility argument. A grid built on variable renewables needs something to cover residual load: the demand remaining after wind and solar have contributed. That residual is small in energy terms and enormous in capacity terms, and it appears at inconvenient times — still winter evenings, multi-day weather systems, the tail of the distribution that storage sized for a normal day does not reach. An open-cycle gas turbine covers exactly that shape: cheap to own, expensive to run, fast to start, and used rarely. Section 10.10’s screening curve says a plant expected to run 3% of the year should be a peaker.

The risk in the bridge argument is duration. Gas plant is built with a 25 to 30 year cost recovery, and a bridge financed over thirty years is a destination for anyone making decisions on a fifteen-year view. The honest version of the argument specifies the capacity factor the plant is expected to run at over its life, and prices the asset accordingly — which almost no financing case does.

10.21 — India and the coal question

India generates roughly three-quarters of its electricity from coal and has the world’s fastest-growing electricity demand. Both facts have to be held at once, and the arithmetic of this chapter explains why the transition looks different here than in systems with flat or falling demand.

• •Growth changes the question. Where demand is flat, adding renewables displaces coal generation directly. Where demand is growing several percent a year, renewables can be added at record pace and coal generation can still rise in absolute terms. Both headlines are true simultaneously and describe the same system.
• •The fleet is young. A large share of Indian coal capacity was built after 2010, so the stranded-asset arithmetic of Section 10.19 has decades of accounting life still to run — a very different position from a fifty-year-old American or European unit.
• •Domestic coal is cheap and imported gas is not. The coal-to-gas switch of Section 10.18 assumes a gas price that Indian generators largely do not see. India has substantial gas capacity running at low utilisation because the fuel is too expensive to dispatch, which is the switching calculation running in reverse.
• •Flexibility is the near-term product. With solar being added at very large scale, the binding problem is increasingly the evening ramp rather than midday energy. That makes coal plant flexibilisation — lowering minimum load, improving ramp rates — and pumped storage and batteries the pressing investments, rather than a straight fuel substitution.
• •Air quality is a separate and more urgent lever. Section 10.14’s controls deliver health benefits that do not depend on climate policy at all, and enforcement of flue gas desulphurisation and particulate standards on the existing fleet is a larger near-term public health intervention than any plausible change to the generation mix.

10.22 — The deeper lesson

Combustion is the only technology in this part of the guide whose fate is decided by a number it does not control. Solar, wind, nuclear and hydro spend their money up front and then run; their risk is construction cost, interest rates and, for the variable ones, when the resource shows up. A coal or gas plant spends comparatively little up front and then buys its output, one gigajoule at a time, in a commodity market that can double in a year.

That single structural fact generates everything in this chapter. It puts combustion at the top of the merit order and therefore in charge of the price. It makes the screening curve the right tool for deciding what to build, and the capacity factor — not the cost per kilowatt — the variable that decides which technology wins. It means a fleet can be rendered uneconomic without any of its costs changing, purely by losing hours. And it makes fuel switching, uniquely, an emissions lever that operates through dispatch rather than construction, which is why it is the fastest one available and the most sensitive to prices nobody in the power sector sets.

• •Efficiency basis — higher or lower heating value? Roughly 10% apart for gas, and most cross-country arguments are secretly this.
• •Heat rate — new and clean at full load, or as the plant actually runs? Part load and fouling both make it worse.
• •Fuel price assumed — every marginal cost, levelised cost and switching price in this chapter hangs off a delivered $/GJ.
• •Capacity factor assumed — above or below the screening crossover? This, not cost per kilowatt, decides which technology is correct.
• •Cost of capital — sets the intercept of every screening line. It punishes capital-heavy plant and barely touches a peaker.
• •Emissions boundary — stack only, or upstream included? Worth about a quarter of a CCGT’s carbon intensity.
• •Warming horizon — GWP over a hundred years or twenty? It changes the coal-versus-gas comparison materially.
• •Cycling regime — starts per year against the design assumption. This is the maintenance cost that never appears in a levelised cost figure.

That closes Part 2. Every technology in it answers the same question — how is a megawatt-hour made, and what does the next one cost? Chapter 11 opens Part 3 by changing the question entirely. Storage makes nothing. It moves electricity that already exists from a moment when it is abundant to a moment when it is valuable, and in doing so it turns the variable resources of Chapters 6 and 7 into something a grid operator can dispatch.