How does the grid stay balanced every second?
Trip a plant and watch the grid catch it
- Foundations · L1
- Lesson 3 of 14
- 15 min
- After lesson 0
- Completed
One plant trips: what holds the frequency
- Inertia: every spinning machine slows together and gives up stored energy (the first seconds).
- Primary response: governors on plants holding reserve add power and stop the fall (seconds).
- Secondary response: automatic generation control raises reserve plants until 60 Hz returns (minutes).
- Last line: if the fall reaches 59.3 Hz, relays disconnect a block of customers automatically.
With JavaScript on, you trip the plant yourself, set the inertia and the reserve, and step through a real night in Texas.
Model. The Plains Grid is illustrative: a 4,000 MW region with one frequency and no power lines.
Time scales. The layers follow NERC (2021); real responses overlap.
Texas. The night of 15 February 2021 is real: readings from the FERC–NERC report (November 2021, pp. 135–138).
What you just did, as diagrams
- Frequency
- The grid's pulse, 60 Hz in North America; it falls whenever supply is short of demand.
- Inertia
- Energy stored in spinning machines; it slows the fall but cannot stop it.
- Primary response
- Governors add power within seconds and stop the fall at its lowest point, the nadir.
- Secondary response
- Automatic generation control brings frequency back to 60 Hz within minutes.
- Reserve
- Spare capacity held back for a trip. In this model, governors and AGC share it on running plants; real reserves can also come from offline plants, storage and responsive loads.
- Automatic load shedding
- The last line: relays cut a block of customers at 59.3 Hz to save the rest of the grid.
Go deeper · AnalystThe first second and where the fall stops, in two lines
Let \(f_0 = 60\) Hz, \(\Delta P\) the lost power in MW and \(E\) the stored rotational energy in MW·s (1 GW·s = 1,000 MW·s). In the first instant only inertia acts, so the rate of change of frequency is
\[\frac{df}{dt}\Big|_{t=0} = -\frac{f_0\,\Delta P}{2E} = -\frac{60 \times 100}{2 \times 16{,}000} \approx -0.19 \text{ Hz/s}.\]
Halve \(E\) and the fall is twice as fast. Once governors have caught up, the frequency settles where the governors’ extra output, \(K\) MW per Hz of drop, and the load’s own relief, \(D\) MW per Hz, cover the lost power. The size of that drop, in Hz below 60, is
\[\text{drop} = \frac{\Delta P}{K + D} = \frac{100}{600 + 80} \approx 0.15 \text{ Hz}, \quad f \approx 59.85 \text{ Hz}.\]
\(K = 600\) MW/Hz is what a 5% droop setting gives on 1,800 MW of plants: a 5% drop in frequency (3 Hz) would call for their full output. All Plains Grid numbers are illustrative.
Go deeper · ModelerWhat the trace's model assumes
- One frequency. The whole region shares it; there are no power lines.
- Stored spin. It sets how fast the frequency moves when power is short.
- Governors. They add power in step with the drop, about 8 s behind, never more than the reserve.
- AGC. It keeps adding, slowly, until 60 Hz returns, also within the reserve.
- Load. Demand eases a little as frequency falls; at 59.3 Hz relays cut one block of 5% of load.
Every 10 ms the page applies one balance:
\[\text{change in frequency} \;\propto\; \frac{\text{power added} - \text{power lost}}{\text{stored energy}}\]
Scheduling and planning models usually skip these seconds and keep a rule instead, such as “hold reserve at least as large as the biggest plant”. Lesson 10 shows how to write a rule like that down, and lesson 11 how a solver meets it.
Go deeperHow big is inertia on a real grid?
- On the night of 15 February 2021, ERCOT's measured inertia ranged from 254 to 349 GW·s; its operators used 100 GW·s as a minimum (FERC–NERC report, 2021, p. 138).
- That is far more stored rotational energy than the Plains Grid's 6–30 GW·s; the amount depends on which machines are connected at the time.
Go deeperWhat the model leaves out
- One frequency for the whole region: on a real grid, machines in different places swing against each other along the power lines.
- Governors have dead bands and limits, and many plants run with no spare capacity at all.
- Real load-shedding programmes cut customers in several smaller steps.
- Real AGC also watches the power flowing to neighbouring regions, not only the frequency.
- Voltage, which can collapse on its own, is not modelled; Analyst lesson A17 makes first estimates of steady voltages and line loadings with short formulas, not a power flow.
Check yourself
Myth or fact? “After a plant trips, governors bring the frequency back to exactly 60 Hz.”
- Myth
- Fact
Answer
Myth. Governors stop the fall but leave the frequency a little low; automatic generation control restores 60 Hz over minutes.
Myth or fact? “With less inertia online, the frequency falls faster after the same trip.”
- Myth
- Fact
Answer
Fact. The first-second fall rate is proportional to the lost power divided by the stored energy: half the inertia, twice the speed.
Myth or fact? “The February 2021 Texas outages were caused by frozen wind turbines.”
- Myth
- Fact
Answer
Myth. Every kind of plant failed. Of the 1,045 units with outages across ERCOT, SPP and MISO South, 58% were gas-fired, 27% wind and 6% coal: shares of units, not of lost MW (FERC–NERC report, 2021, p. 15).
A 200 MW plant trips on a grid storing 20 GW·s (20,000 MW·s) of rotational energy. How fast does the frequency fall in the first instant?
- 0.03 Hz per second
- 0.3 Hz per second
- 3 Hz per second
- 0.6 Hz per second
Answer
0.3 Hz per second: 60 × 200 ÷ (2 × 20,000) = 0.3. Twice the lost power or half the inertia doubles it.
Which of these acts last after a plant trips?
- Inertia
- Governors
- Automatic generation control
- Operators starting another plant to rebuild the reserve
Answer
Operators rebuilding the reserve: this tertiary layer takes ten minutes to hours, so the grid is ready for the next trip.
Numbers, sources and what’s next
Numbers to remember
| Number | What it means | Source |
|---|---|---|
| 0–12 s · 1–10 min | Typical time scales of inertia and of automatic generation control after a disturbance; governors overlap with both, and operators follow. | NERC, Balancing and Frequency Control (PDF), Table 1.1 · May 2021 |
| 59.3 Hz | Where the first block of automatic load shedding starts in Texas (ERCOT, at least 5% of load) and in the Southwest Power Pool (at least 10%). | FERC–NERC report (PDF), pp. 44–45 · settings as of February 2021 |
| 59.304 Hz | ERCOT's lowest frequency on 15 February 2021; it stayed below 59.4 Hz for 4 min 23 s of the 9 minutes its generator protection allowed. | FERC–NERC report (PDF), p. 137 · 15 Feb 2021 |
| 20,000 MW | Customer demand ERCOT ordered off at the worst point of Winter Storm Uri. | FERC–NERC report (PDF), p. 9 · Feb 2021 |
| 4.5 million | Customers in Texas who lost power during Winter Storm Uri. | EIA, Today in Energy · November 2022 |
Next on the starter path
In the real world
Across the region. In ERCOT, SPP and MISO South, 1,045 generating units had 4,124 outages, derates or failures to start; three-quarters came from freezing or fuel problems.
In Texas. Plants failed faster than the reserve could be rebuilt, so operators cut customers by hand to hold the frequency.
Next. Lesson 4 looks at the forecasts that tell operators how much reserve to hold.
Sources and further reading
- NERC Resources Subcommittee (2021). Balancing and Frequency Control Reference Document (PDF). The control continuum, inertia, governors, AGC.
- FERC, NERC and Regional Entity Staff (2021). The February 2021 Cold Weather Outages in Texas and the South Central United States (PDF). Chronology of 15 February, pp. 135–138 (relay trips at 1:54 a.m., p. 137); load-shedding settings, pp. 44–45; unit outages in ERCOT, SPP and MISO South, p. 15.
- EIA (2022). U.S. electricity customers averaged seven hours of power interruptions in 2021. Today in Energy.
Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.
Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.