How the Power Grid Works: Generation, Transmission, Balancing

The electricity grid is the largest machine humans have built, and it has one odd property: it stores almost nothing. Every second, the power fed in by thousands of generators must equal the power drawn by millions of devices plus the losses in the wires between them. When the two drift apart, the frequency moves and operators have seconds to minutes to correct it. This guide covers how the grid is built, how it stays balanced, why some energy is lost on the way, where prices and markets come in, and what changes as wind, solar and batteries take over from large spinning power stations.

The four layers: generation, transmission, distribution, consumption

A conventional grid is a hierarchy of voltage levels. Power stations generate at medium voltage, step-up transformers raise it for long-distance transport, and a chain of substations brings it back down until it reaches your socket.

Why so many voltages? Resistive loss in a line grows with the square of the current: P_loss = I²R. For the same power, doubling the voltage halves the current and cuts the loss to a quarter. So bulk power travels at hundreds of kilovolts, and only the last few hundred metres to a house run at a voltage safe enough to wire into walls.

Worked example: why 400 kV beats 20 kV

Send 500 MW down a 100 km line with a resistance of 0.03 Ω per km per phase (3 Ω total per phase) at a power factor of 0.95.

• At 400 kV, the current is 500,000,000 / (1.732 × 400,000 × 0.95) ≈ 760 A. Loss = 3 × 760² × 3 ≈ 5.2 MW, about 1% of the power sent.
• At 20 kV, the current would be about 15,200 A. The formula gives a loss of 3 × 15,200² × 3 ≈ 2,080 MW, more than the power you are trying to send. No real conductor could carry that current, so the line simply could not be built.

This is the main reason AC won the “war of the currents”. Transformers make stepping up the voltage cheap and efficient, and large units exceed 99% efficiency.

Frequency: the grid’s heartbeat

Every large synchronous generator on one AC grid spins in lockstep. In Europe, most of Asia, Africa and Australia the rhythm is 50 Hz. In North America, parts of South America, Japan’s western half and Saudi Arabia it is 60 Hz. The Continental European synchronous area alone links more than 20 countries and serves several hundred million people at one common frequency.

Frequency gives a direct reading of the power balance:

• When demand exceeds generation, the shortfall comes out of the kinetic energy stored in the rotating masses of turbines and generators. They slow down and frequency falls.
• When generation exceeds demand, the machines speed up and frequency rises.

That rotating mass is called inertia. It does not fix an imbalance, but it slows how fast frequency changes (the RoCoF, rate of change of frequency) and buys time for the controls to react.

Example: a system with 200 GVA of synchronous machines at H = 4 s loses a 1.5 GW nuclear unit. The initial RoCoF is 50 × 1.5 / (2 × 4 × 200) ≈ 0.047 Hz/s. With no response at all, frequency would take about 4 seconds to fall 0.2 Hz. That is long enough for automatic reserves to start acting.

The layers of balancing reserves

Operators keep capacity in reserve and group it by how fast it acts. The table uses European (ENTSO-E) terms with common North American equivalents.

Reserves are sized against a “reference incident”. In Continental Europe that is the simultaneous loss of 3,000 MW of generation. FCR is shared across the whole synchronous area, which is one reason large interconnected grids ride out faults better than islands do.

Voltage, reactive power and stability

Frequency is the same across the whole system. Voltage is local: every node has its own, and it must stay within roughly ±5–10% of nominal. Operators control it mainly through reactive power, the part of AC power that sloshes back and forth to build magnetic and electric fields in lines, motors and transformers without doing useful work.

• Long, lightly loaded lines generate reactive power and push voltage up.
• Heavily loaded lines and induction motors consume reactive power and pull voltage down.
• Generators, capacitor banks, reactors, STATCOMs and modern inverters inject or absorb reactive power to keep voltage in range.

If voltage support fails in a heavily loaded region, a voltage collapse can cascade even when there is enough generation overall. Several historic blackouts, including parts of the 2003 North American event, combined overloaded lines, too little reactive support and protection systems tripping one after another.

Losses: where the energy goes

Losses are the energy dissipated between the generator and the meter. Every consumer pays for them, directly or through network tariffs.

Technical losses divide roughly into transmission (often 1.5–3%) and distribution (the rest), plus transformer no-load losses that occur even when nothing is drawn. The grid emission factor in our electricity CO₂ calculator reflects average generation; losses add a few percent on top of the energy each unit at your meter took to produce.

HVDC: the long-distance specialist

High-voltage direct current links convert AC to DC, transmit it and convert it back. Converter stations cost hundreds of millions, but DC lines have no reactive power problems and lose less per km. The usual break-even distance against AC is roughly 600–800 km overhead and only 50–100 km for submarine cables, because the capacitance of AC cables makes long undersea AC links impractical. Hence DC for offshore wind farms far from shore, the North Sea interconnectors and China’s ultra-high-voltage corridors (some longer than 3,000 km). HVDC also lets two grids with different frequencies, or grids that are not synchronised, trade power.

Markets: how the grid decides who generates

In liberalised systems, generators bid into markets and the TSO dispatches the cheapest combination that the network can carry. A simplified European timeline:

1. Forward markets (years to weeks ahead): hedging contracts.
2. Day-ahead auction (closes at noon the day before): sets an hourly or 15-minute price for each bidding zone using the merit order. Plants are stacked from lowest to highest marginal cost, and the last unit needed sets the price.
3. Intraday trading: continuous adjustments as wind and solar forecasts improve.
4. Balancing markets: the TSO buys reserve capacity and activates energy in real time. Imbalance prices penalise parties whose schedules were off.

The merit order explains why wind and solar, with near-zero marginal cost, push prices down when they are plentiful, and why prices spike when expensive gas peakers set the margin. Why electricity prices spike tells the full story. To compare the lifetime cost of each technology instead of its marginal cost, try the LCOE calculator.

What changes with wind, solar and batteries

Wind turbines, solar panels and batteries connect through power electronics (inverters), not synchronous machines. That changes the engineering in four ways.

1. Less inherent inertia. Inverter-based resources do not spin in lockstep with the grid, so a grid running on lots of wind and solar at a given moment has less rotating mass. Operators compensate with synchronous condensers (generators with no prime mover), fast frequency response from batteries and “synthetic inertia”.

2. Grid-following vs grid-forming inverters. Most inverters today are grid-following: they measure the existing voltage waveform and inject current in step with it. Grid-forming inverters set voltage and frequency themselves and behave like a voltage source. NREL, AEMO and European TSOs see grid-forming capability as a requirement for systems that run near 100% inverter-based generation.

3. Bidirectional distribution. Distribution networks were designed for power flowing one way. Rooftop PV reverses the flow on sunny afternoons and raises local voltage. DSOs now require inverters to provide volt-var control, cap exports, or both. A home battery sized with our solar battery calculator can soak up the midday surplus before it reaches the street.

4. Variability and forecasting. Wind and solar output follows the weather. Balancing moves from following demand to following net load: demand minus wind and solar. On sunny spring days in California or southern Australia this creates the “duck curve”, with low net load at midday and then a steep evening ramp of many gigawatts within a few hours.

Reliability: measuring how well a grid works

Two distribution-level indicators are the usual yardsticks:

• SAIDI: average minutes of interruption per customer per year.
• SAIFI: average number of interruptions per customer per year.

Well-run European urban networks often record a SAIDI under 20 minutes a year. Many US utilities report one to several hours once major storms are counted, and some developing grids see hundreds of hours. Most interruptions start in distribution (fallen trees, faults on local lines) rather than in a shortage of bulk generation.

See the grid for yourself

Look up the generation mix of any country on our world energy map, dig into one nation’s grid in the country profiles, or compare carbon intensity and renewable share in the rankings. For the units grid engineers use (MW, MWh, MVA, TWh), see the energy units cheat sheet and the energy unit converter.

Figures in this guide are typical ranges drawn from the sources above; your own numbers depend on location, prices and equipment. Use the linked calculators for site-specific estimates.