NovaSolver›Smart Grid Frequency Droop Control Simulator Back
Power Grid Frequency
Smart Grid Frequency Droop Control Simulator
Watch how each generator's droop response and the grid inertia H hold the frequency when the load suddenly changes. Sweep grid region, rated output, droop R, load disturbance, inertia and AGC gain to see whether Δf, RoCoF, nadir and recovery time stay within UCTE limits (±0.2 Hz steady / ±0.8 Hz nadir).
After a load disturbance the grid frequency drops, droop response stabilises it and AGC restores it to the nominal 50/60 Hz. Colour shows |Δf| (green → orange → red).
RoCoF is the rate of change of frequency, H the system inertia (s). ΔPi is each generator's response, proportional to Δf and inversely proportional to its droop R.
Power Grid Frequency Droop Control — Primary Response — 50 Hz / 60 Hz
🙋
The grid is kept exactly at 50 or 60 Hz, right? Who actually controls that? Every time I see a blackout headline I wonder.
🎓
Good question. It is not one person — every generator on the grid coordinates autonomously. The rule is simple: "if frequency drops, increase output; if it rises, cut output". Each turbine's governor does this mechanically, with a slope called the droop R, typically 5%. That is the primary response. With the default 1000 MW load step in this tool you can see it settle at Δf = −0.167 Hz.
🙋
But 0.167 Hz sounds tiny. So why do big blackouts still happen?
🎓
Sharp point. The steady value is not the problem — the transient and the disturbance size are. Look at the RoCoF (rate of change of frequency, Hz/s): default −0.42 Hz/s, but lower H from 4 s to 2 s and it roughly doubles. Once RoCoF exceeds 1 Hz/s, protective relays misfire or the nadir crosses the UFLS threshold (47–49 Hz) and cascades trip the load. The 2003 Northeast blackout (50 million affected) and the 2021 Texas cold snap (4 million households, 26 GW lost) were chains that the primary response could not stop.
🙋
So inertia matters a lot. Is that the same "renewables shrink inertia" problem people keep talking about?
🎓
Exactly. Thermal and nuclear turbines are tens of tonnes of spinning steel, so even when frequency dips their kinetic energy buffers it for a moment — that is the source of H ≈ 4–6 s. Solar and wind are inverter-coupled, so by default zero inertia. RES-heavy grids fall to H ≈ 2 s and RoCoF gets wild. The fixes are synthetic inertia and virtual droop from grid-forming inverters, batteries doing FCR (Hornsdale 100 MW in Australia), and synchronous condensers placed where coal plants used to be.
🙋
I also notice "AGC recovery time" on the dashboard. What is that?
🎓
Primary control's job is to "stop the fall, with a residual Δf"; restoring exactly 50/60 Hz is the job of secondary control, AGC (Automatic Generation Control). It integrates the area control error and dispatches set-point corrections. With the default ΔPL = 1000 MW and AGC = 400 MW, recovery is 75 seconds. It runs at the 30 s – 15 min scale; after 15 min tertiary control (economic dispatch) takes over. ENTSO-E formalises this as FCR / FRR / RR; Japan as LFC / EDC.
Frequently Asked Questions
Droop control is the primary control in which each generator fixes a line between its power P and frequency f with a slope equal to the droop R: when frequency drops, output rises. R = 5% means a 5% drop in frequency causes the unit to deliver an extra 100% of rated output. Across the grid the slopes add in parallel to the composite β = Σ Pn,i/(Ri·fn) [MW/Hz]. For a load step ΔPL the steady-state deviation is Δf = −ΔPL/β. This tool also reports the transient RoCoF and nadir governed by the system inertia H.
H is the ratio of stored rotating kinetic energy to rated power and is around 4–6 s in thermal-dominated grids but as low as ~2 s in inverter-rich grids. Small H produces a steep RoCoF = −ΔPL·fn/(2H·Pn,total) right after the disturbance. The nadir crosses the UCTE limit (±800 mHz) before UFLS and protective relays can act. Halving H here (4 s → 2 s) roughly doubles the RoCoF and triggers the warning verdict. Mitigations include synthetic inertia from grid-forming inverters and synchronous condensers at retired thermal sites.
Primary control is the droop response of this tool. It acts in 0–30 s and arrests the frequency excursion, but a steady Δf remains. Secondary control (AGC: Automatic Generation Control) runs at 30 s – 15 min, integrating the area control error (ACE) to restore the nominal frequency (50/60 Hz). The 'AGC recovery time' shown here is an estimate of that. Tertiary control redispatches over 15 min and beyond. UCTE/ENTSO-E formalises this as FCR / FRR / RR; Japan calls it LFC / EDC.
Synchronous machines (thermal, nuclear, hydro) naturally provide rotating inertia and droop. PV and wind, being inverter-coupled, provide neither by default. The grid-wide β and H therefore drop, enlarging Δf and RoCoF for the same disturbance. The three responses are: (1) synthetic inertia and virtual droop from grid-forming inverters, (2) battery FCR such as Hornsdale (100 MW / 129 MWh) in South Australia, and (3) synchronous condensers installed at retired thermal sites. ENTSO-E plans to set minimum inertia and non-synchronous penetration limits toward 2030.
Real-World Applications
Transmission system operators (TSO): Japan's OCCTO, Europe's ENTSO-E and North America's NERC set droop values, FCR procurement volumes and frequency deviation limits (±0.2 Hz steady, ±0.8 Hz nadir, etc.) per zone. The kind of Δf / RoCoF / nadir estimate produced by this tool is used in the initial grid-impact study that a new power plant must submit for grid-code compliance. Capacity and balancing markets are increasingly monetising FCR and FRR through auctions.
Large batteries and grid-forming inverters: Australia's Hornsdale Power Reserve (100 MW / 129 MWh, Tesla) cut South Australia's frequency-control cost by roughly 90% since it came online in 2017. Sizing it requires choosing how to split droop and synthetic inertia so that RoCoF and nadir stay within bounds for the design contingency — exactly the kind of question this tool sketches. GFM (grid-forming mode) standardisation is currently advancing at AEMO, National Grid ESO and METI.
Microgrids and islanded systems: Small islands, data centres and factory microgrids have extreme low inertia (1–2 s), so RoCoF swings violently. Diesel, gas turbines, batteries and PV must be co-dispatched, with each source's droop, dead band and AGC share tuned with care. Setting Pn = 100 MW, H = 1 s in this tool gives a rough taste of an islanded system's response.
RES integration risk assessment: Events like the 2019 UK Hornsea outage (lightning hit, 1.13 M households out), the 2020 Kyushu curtailment regime and the 2021 Texas cold snap (4 M households out, 26 GW lost) are cited in grid-operation guidelines worldwide as cases where "low inertia plus a large contingency" outran the primary response. Pushing H and ΔPL to extremes in this tool helps build that intuition.
Common Misconceptions and Pitfalls
The first trap is assuming "smaller droop R = smaller Δf = better design". Yes, smaller R raises β and shrinks Δf, but it also makes the units more prone to hunting (oscillation between paralleled generators). When several machines share a bus, very small droops make each one twitch before the others can answer, and the power output oscillates indefinitely. In practice R = 4–8% is the workable range; below ~3% the phase margin gets thin. Grid-impact studies size R by the trade-off between Δf suppression and stability margin.
The second trap is thinking "a narrow dead band makes the response faster". With zero dead band the units indeed react to even tiny excursions, but governor valves are driven continuously and wear out, and sensor offsets push all units to respond together — a common-mode oscillation. Real units carry a 10–30 mHz dead band so they ignore steady-operation noise and react only to meaningful disturbances. Setting the dead band to 0 in this tool drives Δf to 0, but the cost is "the machine never stops moving".
The third trap is conflating RoCoF with nadir. RoCoF is the rate of change just after the disturbance (Hz/s) and depends on H; it directly maps to df/dt relay misoperation. Nadir is the lowest frequency reached during the transient (Hz) and directly maps to UFLS thresholds (47–49 Hz, staged trip). For the same ΔPL, lowering H raises RoCoF (relay risk), while lowering β deepens nadir (UFLS risk). The causes differ, so the fixes differ. That is why this tool reports both and why the verdict checks them separately.
How to Use
Set nominal generator output in MW (e.g., 500 MW) and system inertia constant in seconds (e.g., 6 s for large synchronous machine)
Enter droop gain as percentage per Hz (typical 4–6% for primary control) and load disturbance magnitude in MW (e.g., 150 MW step loss)
Run transient sweep to observe frequency nadir, rate-of-change-of-frequency (RoCoF) in Hz/s, generator power injection response, and AGC recovery time to final steady-state frequency
Worked Example
For a 1000 MW grid with 8 s inertia constant, 150 MW load step, and 5% droop: RoCoF peaks at −0.98 Hz/s, nadir drops to 59.62 Hz in 1.2 s, generator injects +75 MW (half the disturbance via 5% droop sensitivity), frequency settles to 59.85 Hz within 45 s as automatic generation control (AGC) brings frequency back to 60 Hz nominal over 300 s. Lower inertia (5 s) steepens RoCoF to −1.47 Hz/s, requiring faster droop response.
Practical Notes
Tighter droop (lower %) increases steady-state frequency error but speeds primary response; 4% droop suitable for wind-heavy systems with low synchronous inertia (<5 s)
RoCoF threshold enforcement (e.g., >1 Hz/s triggers under-frequency load shedding); synchronous condensers or synthetic inertia reduce peak RoCoF by 20–40%
AGC recovery time extends with congested transmission; verify settling frequency matches economic dispatch targets to avoid persistent reserve deployment