Where the electricity comes from
According to Ember's Global Electricity Review 2026, published in April 2026, renewable sources generated 33.8% of the world's electricity in 2025, overtaking coal (33.0%) for the first time. Lumenary EE's renewables and storage tells that story. But read the figures the other way round: the remaining 66.2% came from coal, gas, nuclear and oil, every one of them a heat engine turning a generator. Add the bioenergy, geothermal and solar-thermal plants counted among the renewables, and more than two thirds of the world's electricity in 2025 was still made by heating a fluid and letting it expand through a turbine.
A single coal or nuclear unit delivers 1 GW or more, through a turbine shaft that can be 50 m long, spinning at 3000 or 3600 revolutions per minute to match a 50 or 60 Hz grid (half that in many nuclear plants). The cycles of Power and refrigeration cycles are the skeleton of these plants. This lesson adds the flesh: what real plants add to the textbook cycles and why, how a combined-cycle plant passes 60% efficiency, how a nuclear plant differs, what cogeneration buys, and how to read the capacity factors and "primary energy" figures of national statistics.
Every feature of a power plant you are about to meet answers one of two questions. How can the working fluid receive its heat at a higher average temperature? And how can it give its heat away at a lower one? Keep those two questions beside the diagrams and almost every pipe will explain itself.
Inside a steam power plant
A steam plant is a Rankine cycle stretched across a site. Water is pumped to high pressure, boiled and superheated in the boiler (in a nuclear plant, a steam generator), expanded through the turbine to drive the generator, condensed in the condenser, and pumped round again. The condenser's heat, often the largest energy flow on the site, goes to a river, the sea or the air through a cooling tower.
Of every 100 units of fuel energy in a typical modern coal unit, roughly 6 to 10 leave the stack as warm flue gas, about 50 go to the cooling water, a few are lost in the generator and transformer, and 5 to 8 run the plant itself (fans, pumps, coal mills, flue-gas scrubbers). About 40 leave as electricity. Every thermal plant has the same shape: one large heat rejection set by the second law, and smaller losses that engineers fight one at a time.
Efficiency, heat rate and capacity factor
Definition— Net plant efficiency and heat rate
The net efficiency of a power plant is the electricity it delivers to the grid, after supplying its own auxiliary loads, divided by the energy of the fuel it burns over the same period: Its heat rate is the reciprocal, quoted as fuel energy per kilowatt-hour: kJ/kWh (in the United States often in Btu/kWh, where 3412 Btu = 1 kWh).
The higher heating value (HHV) of a fuel includes the heat released when the water vapour in the flue gas condenses; the lower heating value (LHV) does not, since most plants send that vapour up the stack. For natural gas the HHV is about 11% larger, so the same gas plant can be "61%" (LHV) in a manufacturer's brochure and "55%" (HHV) in a US government table. Always check which.
Definition— Capacity factor
The capacity factor of a generator over a period is the energy it produced divided by the energy it would have produced running at its full rated (nameplate) power for the whole period. Over a year, .
Efficiency says how well a plant converts energy while it runs; capacity factor says how much of the time it effectively runs. They are independent: a gas turbine kept for the hottest afternoons can be efficient yet have a capacity factor near 10%. Lumenary EE's the power grid explains how the merit order decides which plants run.
The Rankine cycle in practice, computed
In the basic Rankine cycle of Power and refrigeration cycles, feedwater enters the boiler barely warmer than the condenser, so much of the heat is received at low temperature, where by the Carnot argument of The second law and entropy it is worth less. Real plants raise the mean temperature of heat addition in four ways.
- Higher pressure and temperature. Subcritical boilers run at about 16 to 18 MPa; supercritical units run above water's critical pressure of 22.06 MPa, where there is no boiling, and ultra-supercritical units reach 25 to 30 MPa and 600 to 620 °C, limited by the creep strength of the hottest steels (Lumenary Materials' fatigue, creep and corrosion).
- Reheat. After the high-pressure turbine the steam returns to the boiler and is heated again. Its main purpose is to keep the exhaust dry: droplets erode the last blades, and designers like the exhaust quality to stay above about 0.88 to 0.90.
- Regenerative feedwater heating. Steam bled from the turbine at several pressures heats the feedwater with latent heat that would otherwise have gone to the condenser. Real plants have seven to ten heaters; one, the deaerator, is an open heater that also strips dissolved oxygen from the water.
- A lower condenser pressure, set by the temperature of the cooling water or air on site.
The program builds a plant up one feature at a time. The steam enthalpies were looked up beforehand from IAPWS-95 with the open-source CoolProp library; the program does the energy bookkeeping, applying the turbine's isentropic efficiency to each partial expansion, a common simplification.
# A steam power plant, built up one improvement at a time.
# Enthalpies h (kJ/kg) and specific volumes v (m^3/kg) were looked up beforehand from the
# IAPWS-95 formulation for water (with the open-source CoolProp library) and rounded.
# "hs_..." is the enthalpy at the end of an ISENTROPIC expansion to that pressure.
SUB = { # subcritical unit: 16.5 MPa, 540 C, reheat at 4.0 MPa to 540 C
"p1": 16.5, "h1": 3406.4, "hs_rh": 2999.9, "hs_cond_noRH": 1995.9, "h3": 3537.5,
"bleeds": {1.6: (3233.6, 858.5, 0.001159), # p (MPa): (hs from the reheat state, hf, vf)
0.6: (2970.3, 670.4, 0.001101),
0.15: (2687.2, 467.1, 0.001053)},
"hs_cond": 2238.7,
}
USC = { # ultra-supercritical unit: 27 MPa, 600 C, reheat at 5.5 MPa to 620 C
"p1": 27.0, "h1": 3474.9, "hs_rh": 3007.3, "hs_cond_noRH": 1959.0, "h3": 3709.5,
"bleeds": {1.6: (3272.8, 858.5, 0.001159),
0.6: (3001.6, 670.4, 0.001101),
0.15: (2709.8, 467.1, 0.001053)},
"hs_cond": 2256.9,
}
P_COND, HF_COND, HFG_COND, VF_COND = 0.007, 163.4, 2408.3, 0.001008 # condenser: 7 kPa, 39 C
def cycle(d, eta_t=0.88, eta_p=0.80, reheat=True, heaters=()):
"""Cycle efficiency and exhaust quality, per kg of steam through the boiler."""
def pump(v, p_lo, p_hi): # kJ/kg: v dp / efficiency (MPa m^3/kg = 1000 kJ/kg)
return v * (p_hi - p_lo) * 1000 / eta_p
h1 = d["h1"]
if not reheat: # one expansion, straight to the condenser
h_exit = h1 - eta_t * (h1 - d["hs_cond_noRH"])
w_pump = pump(VF_COND, P_COND, d["p1"])
w_net = (h1 - h_exit) - w_pump
return w_net / (h1 - HF_COND - w_pump), (h_exit - HF_COND) / HFG_COND
h2 = h1 - eta_t * (h1 - d["hs_rh"]) # HP turbine exhaust ("cold reheat")
h3 = d["h3"] # after the reheater ("hot reheat")
ps = sorted(heaters, reverse=True) # open heaters, highest pressure first
n = len(ps)
hb = [h3 - eta_t * (h3 - d["bleeds"][p][0]) for p in ps] # bled-steam enthalpies
hf = [d["bleeds"][p][1] for p in ps] # each heater delivers saturated liquid
vf = [d["bleeds"][p][2] for p in ps]
h_exit = h3 - eta_t * (h3 - d["hs_cond"])
# Pumps from the bottom up: condenser -> lowest heater -> ... -> top heater -> boiler.
lows = [(P_COND, HF_COND, VF_COND)] + [(ps[k], hf[k], vf[k]) for k in reversed(range(n))]
highs = list(reversed(ps)) + [d["p1"]]
w_pumps = [pump(v, p_lo, p_hi) for (p_lo, _, v), p_hi in zip(lows, highs)]
h_after_pump = [h + w for (_, h, _), w in zip(lows, w_pumps)]
# Bleed fractions, top heater first: y*h_bleed + (flow - y)*h_water_in = flow*hf.
y, flow = [], 1.0
for k in range(n):
h_in = h_after_pump[n - 1 - k] # water reaching heater k from the pump below it
y.append(flow * (hf[k] - h_in) / (hb[k] - h_in))
flow -= y[k]
# Turbine work: the steam flow shrinks at every bleed point.
w_turb, flow, h_prev = h1 - h2, 1.0, h3
for k in range(n):
w_turb += flow * (h_prev - hb[k])
flow -= y[k]
h_prev = hb[k]
w_turb += flow * (h_prev - h_exit)
# Each pump carries the flow leaving the heater (or condenser) below it.
pump_flows = [1.0 - sum(y)] + [1.0 - sum(y[:k]) for k in reversed(range(n))]
w_pump = sum(f * w for f, w in zip(pump_flows, w_pumps))
q_in = (h1 - h_after_pump[-1]) + (h3 - h2) # boiler plus reheater
return (w_turb - w_pump) / q_in, (h_exit - HF_COND) / HFG_COND
cases = [
("ideal simple cycle, 16.5 MPa / 540 C", SUB, dict(eta_t=1.0, eta_p=1.0, reheat=False)),
("+ real turbine (88%) and pumps (80%)", SUB, dict(reheat=False)),
("+ reheat at 4 MPa to 540 C", SUB, dict()),
("+ one open feedwater heater (0.6 MPa)", SUB, dict(heaters=(0.6,))),
("+ three feedwater heaters", SUB, dict(heaters=(1.6, 0.6, 0.15))),
("ultra-supercritical 27 MPa / 600 / 620 C", USC, dict(heaters=(1.6, 0.6, 0.15))),
]
print(f"{'case':44}{'cycle eff.':>11}{'exhaust quality':>17}")
for name, d, kw in cases:
eta, x = cycle(d, **kw)
print(f"{name:44}{eta:10.1%}{x:15.3f}")
# From cycle efficiency to the plant's net efficiency and heat rate.
for name, d in (("subcritical", SUB), ("ultra-supercritical", USC)):
eta_cycle, _ = cycle(d, heaters=(1.6, 0.6, 0.15))
eta_net = eta_cycle * 0.92 * 0.985 * (1 - 0.07) # boiler, generator, auxiliaries (7%)
print(f"{name:20} net efficiency {eta_net:.1%}, heat rate {3600 / eta_net:,.0f} kJ/kWh")The ideal simple cycle reaches 43.2%, but with a real turbine and pumps it drops to 37.9% and, worse, its exhaust quality is 0.831: 17% of the steam leaving the last blades is water. Reheat lifts the quality to 0.926 and the efficiency to 39.9%. One feedwater heater adds 2.7 points, the next two together only 1.4 more: each heater pays less than the one before, which is why plants stop at seven to ten. Moving to ultra-supercritical steam adds another 2.6 points.
To change the pressures or temperatures you need a property library. With CoolProp on your own computer, the state points come from calls like these:
# Not run here: needs CoolProp installed on your own computer (pip install CoolProp).
from CoolProp.CoolProp import PropsSI
h1 = PropsSI("H", "P", 16.5e6, "T", 540 + 273.15, "Water") / 1e3 # kJ/kg at 16.5 MPa, 540 C
s1 = PropsSI("S", "P", 16.5e6, "T", 540 + 273.15, "Water") / 1e3 # kJ/(kg K)
h2s = PropsSI("H", "P", 4.0e6, "S", s1 * 1e3, "Water") / 1e3 # isentropic end at 4 MPaCombined-cycle gas plants
A gas turbine, a Brayton cycle, receives its heat at a very high temperature: gas enters the first turbine blades at 1500 to 1700 °C in the largest machines of 2026, thanks to single-crystal nickel superalloy blades, ceramic thermal-barrier coatings and internal air cooling. But it also rejects its heat hot, with exhaust at 550 to 650 °C, so on its own it manages about 35 to 44%.
The steam cycle has the opposite profile: it receives heat at a few hundred degrees and rejects it near ambient. In a combined-cycle gas turbine (CCGT) plant the two run in series: the gas turbine's exhaust passes through a heat-recovery steam generator (HRSG), a boiler with no burner, which raises steam for a steam turbine.
If the gas turbine converts a fraction of the fuel energy into work, a fraction leaves in the exhaust. If the bottoming steam cycle converts a fraction of that into work (counting what the HRSG fails to recover as part of its loss), the combined efficiency is A 40% gas turbine with a bottoming cycle that converts a third of its exhaust heat gives .
What limits the bottoming cycle is the pinch point. The exhaust cools steadily, but water boils at a fixed temperature; plotted against the heat transferred, the two temperatures come closest where the gas leaves the evaporator, and that gap must stay at least 5 to 10 K or the heat-transfer area (Heat transfer) becomes enormous.
The program models a large gas turbine as an air-standard Brayton cycle with real component efficiencies (its 1500 °C turbine inlet stands in for the temperature after cooling air has mixed in), then an HRSG with one steam pressure, tried at four values.
# A combined-cycle plant: a gas turbine whose hot exhaust raises steam for a steam turbine.
T0 = 288.15 # K: ambient air, 15 C (the ISO rating condition)
cp_a, k_a = 1.005, 1.40 # air in the compressor: kJ/(kg K), ratio of specific heats
cp_g, k_g = 1.15, 1.333 # hot combustion gas in the turbine
def gas_turbine(r=23.0, T3=1773.0, eta_c=0.88, eta_t=0.90):
"""Brayton cycle per kg of air: net work, heat added, exhaust temperature (K)."""
T2 = T0 * (1 + (r ** ((k_a - 1) / k_a) - 1) / eta_c) # real compressor outlet
T4 = T3 * (1 - eta_t * (1 - r ** (-(k_g - 1) / k_g))) # real turbine outlet
w_net = cp_g * (T3 - T4) - cp_a * (T2 - T0)
q_in = cp_g * (T3 - T2) # heat released by the fuel
return w_net, q_in, T4
w_gt, q_in, T4 = gas_turbine()
print(f"gas turbine alone: {w_gt:.0f} kJ per kg of air, efficiency {w_gt / q_in:.1%}, "
f"exhaust {T4 - 273.15:.0f} C")
# Heat-recovery steam generator (HRSG) with ONE steam pressure; steam superheated to 565 C.
# Water/steam enthalpies (kJ/kg) from IAPWS-95 via CoolProp; condenser at 5 kPa (33 C).
H_COND, V_COND, P_COND, HFG_COND = 137.7, 0.001005, 0.005, 2423.0
STEAM = { # p (MPa): (T_sat in C, h of saturated liquid, h at 565 C, h after isentropic expansion to 5 kPa)
2: (212.4, 908.5, 3612.4, 2321.7),
4: (250.4, 1087.5, 3594.6, 2218.9),
8: (295.0, 1317.3, 3558.1, 2110.8),
12: (324.7, 1491.5, 3520.2, 2042.9),
}
PINCH, ETA_ST, ETA_PUMP = 10.0, 0.88, 0.80
print("\nsteam steam per stack exhaust heat turbine-exit combined cycle")
print("p (MPa) kg of gas (C) recovered quality gross net")
for p, (T_sat, h_f, h_sh, h_s) in STEAM.items():
h_feed = H_COND + V_COND * (p - P_COND) * 1000 / ETA_PUMP
T_pinch_gas = T_sat + 273.15 + PINCH # gas leaving the evaporator
m_s = cp_g * (T4 - T_pinch_gas) / (h_sh - h_f) # superheater + evaporator energy balance
T_stack = T_pinch_gas - m_s * (h_f - h_feed) / cp_g # economiser energy balance
w_st = m_s * (ETA_ST * (h_sh - h_s) - (h_feed - H_COND))
eta_gross = (w_gt + w_st) / q_in
eta_net = eta_gross * 0.985 * (1 - 0.02) # generators, auxiliaries (2%)
recovered = (T4 - T_stack) / (T4 - T0) # share of the exhaust's heat (above ambient) used
x_exit = (h_sh - ETA_ST * (h_sh - h_s) - H_COND) / HFG_COND
print(f"{p:5} {m_s:9.3f} {T_stack - 273.15:6.0f} {recovered:10.0%} {x_exit:12.3f}"
f" {eta_gross:7.1%} {eta_net:6.1%}")The gas turbine alone converts 44.7% of the fuel's energy, with 633 °C exhaust; the steam cycle lifts the plant to 59.6 to 61.2% net. The table shows the designer's dilemma: higher pressure gives more work per kilogram of steam, but water boils hotter, so less heat is recovered (85% at 2 MPa, 80% at 12 MPa), the stack runs hotter (138 °C at 12 MPa against 106 °C at 2 MPa), and the turbine exhaust grows too wet (a quality of 0.859 at 12 MPa).
Large plants escape it with three pressure levels, so that steam boils at three temperatures and follows the cooling gas closely, plus reheat to keep the exhaust dry. With such bottoming cycles and the hottest gas turbines, the best plants report net efficiencies above 63% (LHV); one of the first certified at that level was Chubu Electric's Nishi-Nagoya plant in Japan, in 2018.
Combined-cycle plants dominate new fossil generation because they are efficient, quick to build and able to change output within minutes. Their gas turbines are being adapted to burn hydrogen blends; hydrogen burns hotter and faster, which makes low nitrogen oxides and stable flames harder, and it helps the climate only if it is made with low-carbon energy.
Nuclear power plants in outline
A nuclear plant is a steam plant whose boiler has been replaced by a reactor. Each fission of a uranium-235 nucleus releases about 200 MeV, mostly as the kinetic energy of the fragments, which becomes heat in the fuel. Fissioning one gram of uranium-235 releases about 80 GJ, close to one megawatt-day, against roughly 25 to 30 kJ from a gram of coal: a factor of about three million (compare Lumenary Chemistry's energy chemistry).
More than four in five of the reactors in service are light-water reactors, cooled and moderated by ordinary water.
- In a pressurised-water reactor (PWR), the primary water is kept at about 15.5 MPa so that it does not boil at around 320 °C; it passes through steam generators, large tube bundles that boil the separate secondary water at about 6 to 7 MPa and some 280 °C. Only the secondary steam reaches the turbine.
- In a boiling-water reactor (BWR), water boils in the core itself at about 7 MPa, and the steam goes directly to the turbine.
Either way the steam is barely superheated, because the fuel cladding and pressure vessel limit the coolant temperature. With such a low mean temperature of heat addition, light-water reactors convert only about 33 to 37% of their heat into electricity, through huge turbines with moisture separators and reheaters between stages to keep the steam dry enough for the last blades.
One difference from a furnace dominates reactor safety. When a reactor shuts down, the chain reaction stops within seconds, but the fission products keep decaying and heating the fuel: about 6% of full power at once, about 1% after a few hours. For a reactor making 3,000 MW of heat, 1% is 30 MW, enough to boil away its water and damage the fuel if cooling stops. At Fukushima Daiichi in 2011 the reactors shut down when the earthquake struck, but the tsunami disabled the emergency power and cooling, and decay heat damaged three cores. Hence defence in depth, with several independent barriers and cooling systems, and newer designs with passive cooling that works by gravity and natural circulation without power or operators.
As of mid-2026 the IAEA's PRIS database listed a little over 410 power reactors in operation, about 380 GW, supplying about 9% of the world's electricity, with most new construction in China and projects in India, Russia, Türkiye, Egypt, Bangladesh, the United Kingdom and elsewhere. Small modular reactors (SMRs) of up to a few hundred megawatts aim to be factory-built and quicker to deploy: Russia's floating Akademik Lomonosov has run since 2020, China's gas-cooled HTR-PM since 2023, China's 125 MW Linglong One was completing commissioning in 2026, and a 300 MW SMR has been under construction at Darlington, Canada, since 2025. Whether SMRs will be cheaper per megawatt than large reactors is not yet known. Fusion appears in Engineering for space and fusion.
Cogeneration: using the rejected heat
A condenser rejects about half the fuel's energy at 30 to 40 °C, too cool to use. Condense the steam at 90 to 120 °C instead and the rejected heat can warm a city through a district-heating network or supply a paper mill or refinery. This is cogeneration, or combined heat and power (CHP): common in the district heating of Denmark, Finland, Poland and cities from Seoul to Moscow, in industry worldwide, and in the sugar mills of India, Brazil and Mauritius, which burn cane fibre (bagasse) for their own steam and sell surplus electricity.
CHP is often advertised by its total efficiency, electricity plus useful heat over fuel, which can reach 80 to 90%. The program compares three ways of delivering the same heat and power, then asks what the heat is worth.
import math
# Cogeneration (combined heat and power) compared with making heat and power separately.
# Every option delivers the same 40 MWh of electricity and 45 MWh of heat to a district network.
E, H = 40.0, 45.0 # MWh
options = {
"separate: CCGT (58%) + gas boilers (92%)": E / 0.58 + H / 0.92,
"CHP gas engine or turbine (40% + 45%)": 100.0,
"CCGT (58%) powering heat pumps (COP 3.5)": (E + H / 3.5) / 0.58,
}
print("fuel needed for 40 MWh of electricity and 45 MWh of heat")
for name, fuel in options.items():
print(f" {name:44} {fuel:6.1f} MWh of fuel")
saving = 1 - options["CHP gas engine or turbine (40% + 45%)"] / options["separate: CCGT (58%) + gas boilers (92%)"]
print(f"CHP saves {saving:.1%} of the fuel used by separate production")
# Energy counts every MWh alike; exergy asks how much work each MWh could still give.
T0 = 283.15 # K: surroundings at 10 C
T_supply, T_return = 363.15, 323.15 # K: district heating at 90 C out, 50 C back
T_mean = (T_supply - T_return) / math.log(T_supply / T_return) # mean temperature of the heat
carnot = 1 - T0 / T_mean
print(f"\nheat delivered between 90 and 50 C has a mean temperature of {T_mean - 273.15:.1f} C;")
print(f"each MWh of it is worth at most {carnot:.3f} MWh of work (its exergy)")
print(f"CHP: energy efficiency {(E + H) / 100:.0%}, exergy efficiency {(E + H * carnot) / 100:.1%}")
print("CCGT alone: energy and (almost) exergy efficiency 58%")
# The 'power loss' of taking heat from a steam turbine instead of condensing the steam.
T_cond, T_extract = 303.15, 368.15 # K: condensing at 30 C, or at 95 C to heat the network
loss_ideal = 1 - T_cond / T_extract # work forgone per unit of heat, for a reversible turbine
loss_real = 0.85 * loss_ideal # roughly 85% of it for a real turbine
print(f"\nheat extracted at 95 C costs {loss_real:.3f} MWh of electricity per MWh of heat:")
print(f"like a heat pump with a COP of {1 / loss_real:.1f}")Against a modern CCGT plus gas boilers, CHP saves 15.2% of the fuel: real, but far less than "85% against 58%" suggests. Exergy explains why. Heat delivered between 90 and 50 °C has a mean temperature of 69.6 °C, so each megawatt-hour of it is worth at most 0.174 MWh of work, and the CHP plant's exergy efficiency is 47.8%, below a CCGT's 58%. A CCGT driving heat pumps (COP 3.5) needs only 91.1 MWh of fuel, less than CHP, before grid losses.
The last two lines are the reason large steam CHP plants are still worth building. Taking heat out of a steam turbine at 95 °C instead of condensing it at 30 °C costs only about 0.15 MWh of electricity for each megawatt-hour of heat, so the turbine is behaving like a heat pump with a COP of 6.7, better than most real heat pumps. Cogeneration is excellent when the heat is wanted at a modest temperature, near the plant, for many hours a year.
District heating and cooling, and heat pumps in buildings, are the subject of Heating, cooling and buildings.
Capacity factor and efficiency in national energy systems
A country's electricity system is a fleet of very different machines, and two numbers describe each one: how efficiently it converts its fuel, and how much of the year it effectively runs. The US Energy Information Administration publishes capacity factors by technology every month in its Electric Power Monthly; the figure shows the annual values for 2025 from its October 2026 release.
Nuclear plants have low fuel costs and run flat out, stopping mainly to refuel every 18 to 24 months. Wind and solar follow the weather and the sun. Coal and gas plants run when the market needs them, and peakers, simple-cycle gas turbines, are built for a few hundred to a thousand hours a year. Hydropower depends on rainfall and is often saved for when it is most valuable. Capacity factors also differ by country: France's nuclear fleet follows load and runs at lower factors than the American one, and offshore wind farms reach higher ones than those on land.
The second program turns capacity factors into what a planner needs, and then shows a trap in reading national statistics.
# Capacity factors of US utility-scale generators in 2025, from the US Energy Information
# Administration's Electric Power Monthly (tables 6.07.A and 6.07.B, October 2026 release).
CF_2025 = {
"nuclear": 0.911, "geothermal": 0.659, "gas combined cycle": 0.587, "coal": 0.488,
"hydropower": 0.344, "wind": 0.338, "solar PV": 0.240, "gas turbine (peaker)": 0.136,
}
HOURS = 8760
print("source GWh per MW per year MW needed for 1 TWh a year")
for source, cf in CF_2025.items():
print(f"{source:24}{cf * HOURS / 1000:12.2f}{1e6 / (cf * HOURS):20.0f}")
# Three ways to count a country's energy. An invented country generates 200 TWh of electricity.
gen = {"coal": 40, "gas": 30, "nuclear": 60, "hydro": 20, "wind": 30, "solar": 20} # TWh
fossil_eff = {"coal": 0.38, "gas": 0.55}
# (1) Physical energy content (the IEA's convention): nuclear counts as the heat its reactors
# made, at an assumed 33% efficiency; hydro, wind and solar count as the electricity itself.
physical = {s: g / fossil_eff[s] if s in fossil_eff else g / 0.33 if s == "nuclear" else g
for s, g in gen.items()}
# (2) Substitution (the Energy Institute's Statistical Review): non-fossil electricity counts as
# the fuel a typical thermal plant (here 40% efficient) would have burned to make it.
substituted = {s: g / fossil_eff.get(s, 0.40) for s, g in gen.items()}
print("\nsource share of share of 'primary energy'")
print(" electricity physical content substitution")
for s in gen:
print(f"{s:10}{gen[s] / sum(gen.values()):9.0%}{physical[s] / sum(physical.values()):15.0%}"
f"{substituted[s] / sum(substituted.values()):14.0%}")
print(f"total primary energy (TWh): {sum(physical.values()):.0f} by physical content, "
f"{sum(substituted.values()):.0f} by substitution")To produce 1 TWh a year takes 125 MW of nuclear capacity at the 2025 US capacity factor, 338 MW of wind, 476 MW of solar or 839 MW of peaking gas turbines. Shares of capacity and shares of generation are therefore very different things, and headlines confuse them constantly.
The second table shows a subtler trap. To add electricity to coal, oil and gas in a total of primary energy, statisticians must decide what a megawatt-hour of non-fossil electricity counts as. The IEA's physical energy content method counts nuclear as its reactor heat (at an assumed 33% efficiency) and hydro, wind and solar as the electricity itself. The Energy Institute's Statistical Review of World Energy uses substitution, counting non-fossil electricity as the fossil fuel it replaced. In the invented country, wind is 15% of the electricity but 7% of primary energy by the first method and 15% by the second; nuclear is 30% of the electricity and 44% by the first. Neither is wrong, but mixing publications misleads.
Under the physical-content convention, replacing a coal plant with a wind farm cuts "primary energy" by more than half for the same electricity. An electrified, renewable system looks as if it needs far less energy, partly because it really is more efficient and partly because of how it is counted. Read the conventions in the methodology notes before comparing countries or decades.
Ember's data explorer and the IEA's country pages give dated generation by source for every country. As of the mid-2020s, France got most of its electricity from nuclear, Norway nearly all of its from hydropower, Kenya a large share from geothermal steam, India, Indonesia and South Africa most of theirs from coal, and many Gulf states nearly all of theirs from gas and oil. Each mix brings its own engineering problems: cooling water in hot, dry summers, flexible coal fleets as solar grows, or reservoirs in drought years.
How it is done in industry
Plants are designed with heat-balance software that does what the programs above do, for every component and with accurate properties: commercial packages such as Thermoflow's GT PRO and STEAM PRO, Ebsilon Professional and Aspen Plus, and open-source tools such as the Python library TESPy and the process simulator DWSIM. Component designers then use computational fluid dynamics for blades and combustors, heat-transfer correlations for boilers and condensers, and finite-element analysis for creep and thermal fatigue in thick headers and rotors (Failure and fatigue).
Boilers, HRSGs and their pipework are pressure equipment, and a failure can kill, so they are built under codes such as the ASME Boiler and Pressure Vessel Code or the European standards under the EU's Pressure Equipment Directive, and inspected independently throughout their lives. Guaranteed performance is measured under test codes such as ASME's Performance Test Codes.
In 2026 the priorities come from the energy transition. Plants designed to run flat out must now cycle, starting daily and ramping to balance wind and solar, which costs efficiency and causes thermal fatigue. Some coal plants are converting to biomass, gas turbines are being adapted for hydrogen, and a few plants capture their carbon dioxide (SaskPower's Boundary Dam unit in Canada has done so since 2014). Nuclear operators are extending reactor lives to 60 years and in some cases 80. Generators and the grid connection are in Lumenary EE's transformers and machines.
A short history
Thomas Edison's Pearl Street Station in New York (1882) generated electricity with reciprocating steam engines and soon sold its exhaust steam to heat nearby buildings, so the first commercial power station was also an early cogeneration plant. In 1884 Charles Parsons in England built the first practical steam turbine, and by the 1900s turbines had replaced piston engines in power stations. Boiler explosions multiplied too; after disasters such as the 1905 explosion at the Grover shoe factory in Brockton, Massachusetts, which killed 58 people, ASME published its first boiler code in 1914–1915. Brown Boveri commissioned the first gas turbine built for a power station in Neuchâtel, Switzerland, in 1939. The first nuclear power plant connected to a grid was at Obninsk in the Soviet Union in 1954, followed by Calder Hall in England (1956) and Shippingport in the United States (1957). Combined-cycle plants spread from the 1970s and dominated new thermal plant from the 1990s. Each step raised the temperature at which heat is received, or made use of the heat rejected.
Common misconceptions
Misconception
"Power plants waste most of their energy because engineers have not tried hard enough." The largest loss, the condenser's heat, is required by the second law. Engineers have raised efficiency from under 10% in 1900 to over 60% by raising source temperatures and stacking cycles, but no heat engine can reach 100%.
Misconception
"A gigawatt of solar is equivalent to a gigawatt of nuclear." They have the same rated power but, at 2025 US capacity factors, the nuclear plant produces nearly four times as much energy in a year, and at times of the plant's choosing rather than the sun's. Compare energy (TWh) and the times it is available, not capacity (GW).
Misconception
"CHP is 85% efficient, so it beats a 60% combined-cycle plant." Adding megawatt-hours of warm water to megawatt-hours of electricity adds energies of very different quality. Measured by fuel saved against the best alternatives, or by exergy, cogeneration's advantage is real but much smaller, and heat pumps can beat it.
Misconception
"A shut-down reactor is cold and safe." Decay heat continues for years: about 6% of full power at shutdown and around 1% after a few hours. Removing it reliably is the central task of reactor safety systems, and spent fuel needs cooling for years before dry storage.
Exercises
Use 1 kWh = 3600 kJ and 8760 hours in a year. Two exercises ask you to write Python.
1. Power plants and energy conversion
Warm-upHumanA combined-cycle plant has a net efficiency of 58% (LHV). What is its heat rate, in kJ of fuel energy (LHV) per kWh of electricity, to the nearest whole number?
2. Power plants and energy conversion
Warm-upHumanA nuclear unit rated at 1,100 MW (net) delivered 8.9 TWh to the grid in a year of 8760 hours. What was its capacity factor, as a whole-number percentage?
3. Power plants and energy conversion
CoreHumanA gas turbine converts 41% of its fuel's energy into work. Its exhaust passes through a heat-recovery steam generator and a steam cycle which, together, turn 34% of the exhaust's energy into work. What is the combined efficiency, as a whole-number percentage?
4. Power plants and energy conversion
CoreHumanWhich of these changes would lower the cycle efficiency of a steam power plant?
5. Power plants and energy conversion
CoreHumanWhich of these statements are true? Select all that apply.
6. Power plants and energy conversion
CoreHumanA classic rough fit for the decay heat of a reactor that has run for a long time is , where is the time since shutdown in seconds and the thermal power before shutdown (it is reasonable from about ten seconds to a few months). A reactor produced 3,000 MW of heat before it shut down. Estimate its decay heat one day later, in MW, to the nearest whole number.
7. Power plants and energy conversion
CoreComputationalWrite hrsg(T_gas_in, T_sat, h_f, h_sh, h_feed, pinch=10.0, cp_g=1.15) for a single-pressure heat-recovery steam generator, as in the lesson. Temperatures are in °C, enthalpies in kJ/kg and cp_g in kJ/(kg K). The exhaust gas enters at T_gas_in and leaves the evaporator pinch kelvin hotter than the boiling water (T_sat). Steam leaves the superheater with enthalpy h_sh, water leaves the economiser as saturated liquid (h_f), and feedwater enters the economiser with enthalpy h_feed.
Return (m_s, T_stack): the kilograms of steam raised per kilogram of gas, and the stack temperature in °C.
8. Power plants and energy conversion
StretchComputationalUse the lesson's gas-turbine model (ambient 288.15 K; air kJ/(kg K), ; hot gas kJ/(kg K), ).
- Write
gas_turbine(r, T3, eta_c=0.88, eta_t=0.90)returning(efficiency, w_net, T4): the cycle efficiency, the net work in kJ per kg of air, and the exhaust temperature in K. - Write
best_ratios(T3)returning(r_eff, r_work): the integer pressure ratios from 2 to 60 that give the highest efficiency and the highest net work.
Then try best_ratios(1400.0) and compare the exhaust temperatures at the two ratios. Which would you choose for a simple-cycle machine, and which for one feeding a steam cycle?
Further reading and resources
- Electricity 2026 · International Energy Agency · 2026The IEA's annual analysis and forecast of electricity demand and supply, published in February 2026, now looking five years ahead.
- Global Electricity Review 2026 · Ember · 2026Electricity generation by source for the world and for individual countries in 2025, published April 2026, with downloadable data.
- U.S. Energy Information Administration: Electric Power MonthlyMonthly US data, including capacity factors (tables 6.07.A and 6.07.B) and average heat rates by technology.
- Statistical Review of World Energy · Energy InstituteAnnual world energy data by country and fuel, with primary energy counted by the substitution method; its methodology notes explain the conventions.
- IAEA Power Reactor Information System (PRIS)The International Atomic Energy Agency's database of every power reactor in operation, under construction or shut down, with performance data.
- World Nuclear Association information libraryExplanations of reactor types, the fuel cycle and nuclear power country by country.
- The International Association for the Properties of Water and Steam (IAPWS)Publishes the formulations for water and steam that every steam table and property library is built on.
- CoolPropOpen-source thermophysical properties for water and more than a hundred other fluids; used to prepare the steam data in this lesson.
- ASMEPublishes the Boiler and Pressure Vessel Code, born of a wave of boiler explosions, and the Performance Test Codes used to measure plant output and efficiency.