When a generator trips off the grid — a 600 MW nuclear unit at full output, suddenly disconnected — the effect propagates through the interconnection in milliseconds. Before any human operator can read a screen, before any software dispatch algorithm processes the event, before any automatic generation control (AGC) signal has been computed and transmitted, the grid's frequency has already begun to fall. The inertia of rotating synchronous generators is what buys time, but that time is measured in seconds, not minutes. The decisions that matter happen at a speed that most grid software discussions barely acknowledge.
Understanding why 200 milliseconds is the critical response window — not 30 seconds, not 5 minutes, not the 10-minute NERC BAL-002 emergency response timeline — requires understanding the physics of frequency deviation and what happens when the grid can't recover from it fast enough.
Frequency Deviation: The Physical Sequence
North American grids operate at 60 Hz under normal conditions. This is not merely a convention — the synchronous speed of rotating generators is tied to system frequency, and deviations from 60 Hz represent an imbalance between real power generation and real power load. When load exceeds generation, frequency falls. When generation exceeds load, frequency rises.
NERC's frequency response standard (BAL-003-1) establishes that interconnection frequency must not fall below 59.95 Hz (the C-1 frequency response threshold) under normal contingency conditions, and must not reach 59.5 Hz (the threshold for the first stage of automatic under-frequency load shedding, UFLS) under design basis contingencies. The span between 60.0 Hz and 59.5 Hz — a 500 millihertz deviation — is the operating margin that the entire primary frequency response system must protect.
The frequency deviation curve after a large generation trip has a characteristic shape. The rate of change of frequency (RoCoF) immediately after the trip determines how quickly the system approaches those thresholds. RoCoF is inversely proportional to system inertia: a high-inertia grid (many large synchronous generators online) has a slow initial frequency decline; a low-inertia grid (high penetration of inverter-based renewables, which don't inherently contribute rotational inertia) has a fast initial decline. As renewable penetration has increased across ERCOT, CAISO, and the British grid, system operators have observed and documented rising RoCoF values during contingency events — a trend that makes the speed of primary response more important, not less.
Primary Frequency Response: The 30-Second Window
Primary frequency response is the first line of automated grid defense. In synchronous generators, it's implemented through the governor — a mechanical or electronic control system that increases steam or gas input to the turbine when frequency drops, proportional to the frequency deviation. A governor with a 5% droop characteristic will increase output by 5% of rated capacity for a 0.15 Hz frequency drop (from 60.0 to 59.85 Hz). This response is autonomous — it requires no communication with the ISO, no operator action, no AGC signal. The generator senses frequency through its shaft speed and responds within a few seconds.
FERC Order 842, issued in 2018, required all new generators above 1 MW to be capable of and configured to provide primary frequency response. The standard requires full response capability within 10 seconds of a frequency deviation event, with response sustained for 10 minutes. Most governor systems actually reach full response faster — 3–6 seconds — but the 10-second standard represents the regulatory minimum.
Here's the gap: 10 seconds of governor response time is fast for a large steam turbine, but inadequate for the scenarios where frequency recovery depends on non-synchronous resources. Battery storage systems, demand response assets, and HVAC curtailment all have the physical capability to respond in under 1 second — but only if they have frequency measurement capability and automated dispatch logic that triggers at the appropriate frequency threshold without waiting for an ISO dispatch signal.
Why 200 Milliseconds Specifically
The 200 ms figure is not arbitrary. It represents the practical boundary between resources that can influence the frequency nadir — the lowest point the frequency reaches before recovery — and resources that respond after the nadir has already been determined.
The frequency nadir in a modern North American grid contingency event typically occurs 8–20 seconds after the triggering event, depending on the size of the generation loss and system inertia. Resources that respond within the first 1–2 seconds can meaningfully reduce the depth of the nadir. Resources that respond in 5–10 seconds still contribute to the recovery trajectory. Resources that take 30 seconds or longer are providing secondary response — they help restore frequency to nominal but have no influence on whether the nadir breached critical thresholds.
For a grid-connected battery storage system or a commercial building load control system, 200 ms is achievable and represents a meaningful differentiation in grid services value. The measurement chain is: grid voltage/frequency sensor → edge compute → decision logic → control output → load actuation. Each stage has a latency budget:
- Frequency measurement: Phasor measurement units (PMUs) report at 30–120 samples per second; a 60-sample-per-second PMU provides a new frequency measurement every 16.7 ms. Basic frequency measurement via phase-locked loop (PLL) in a commercial inverter is typically 20–50 ms.
- Threshold detection and decision: Sub-50 ms for a purpose-built edge controller running dedicated firmware; 100–500 ms for a general-purpose industrial PC running a polling-based SCADA application.
- Control output and actuation: BACnet write command to a commercial AHU controller typically completes in 80–150 ms on a well-configured Ethernet segment. A relay contact or direct digital output fires in under 10 ms.
A system architected with dedicated edge hardware — not a general-purpose BMS polling loop — can achieve end-to-end response from frequency threshold breach to load reduction in 150–250 ms. A system relying on a standard BMS polling architecture typically achieves 500 ms–2 seconds, which is still valuable for primary response contribution but misses the nadir-influencing window.
ERCOT's Fast Frequency Response Requirement
ERCOT provides the clearest example of a grid operator that has formalized sub-second response requirements. ERCOT's Fast Frequency Response (FFR) product, established after extensive low-inertia analysis following the 2011 Southwest cold weather event, requires qualifying resources to provide full output within 300 ms of detecting a frequency deviation below 59.7 Hz. FFR resources are compensated through ERCOT's ancillary services market, with FFR clearing prices typically in the range of $3–$15/MW-hour for the reservation capacity.
FFR-capable resources in ERCOT include battery storage systems, demand response assets, and HVAC-based frequency-responsive load. The key design requirement is autonomous frequency sensing — the resource cannot wait for an ERCOT dispatch signal, which arrives via the AGC telemetry link on a 4-second scan rate. The resource must detect the frequency event locally and respond immediately, with the ERCOT control room verifying response via real-time telemetry after the fact.
This autonomy requirement is where most commercial building control systems fall short. A building BESS managed by an energy management system that queries a central server for dispatch instructions cannot provide FFR. The decision logic must reside at the inverter or at a local edge controller with direct frequency sensing capability.
FERC Order 2842 and the Inverter-Based Resource Discussion
FERC's ongoing proceedings on inverter-based resource (IBR) grid reliability — including the technical conference on primary frequency response requirements and the Order 2842 proceeding on essential reliability services — reflect regulatory acknowledgment that as inverter-connected resources (solar, wind, storage) displace synchronous generation, primary frequency response must come from sources that can provide it at the required speed without natural inertia.
Grid-forming inverter technology addresses part of this problem by allowing inverters to synthesize inertia — providing a voltage and frequency reference rather than following one. Grid-forming mode is now appearing in commercial BESS products and large solar-plus-storage installations. It does not restore the rotational inertia that a synchronous generator provides, but it does allow inverter-based resources to provide fast frequency response that mimics the governor response profile of synchronous machines, without the mechanical lag.
We're not saying that inverter-based resources can fully replace the frequency response characteristics of large synchronous generators in all scenarios — the physics of virtual inertia and real rotational inertia are genuinely different, and there are transient stability scenarios where the distinction matters. What we're saying is that the 200 ms response capability of well-designed inverter-based resources and automated demand response systems is the design target that the grid needs commercial resources to meet, and that achieving it is an engineering problem with known solutions, not a fundamental barrier.
What This Means for Commercial Building Control
For a building energy engineer or portfolio operator, the practical takeaway from the 200 ms discussion is architectural. Two different control designs lead to fundamentally different grid services capability:
Architecture A — BMS-centric polling: Central BMS (Tridium N4 or equivalent) polls building equipment every 15–30 seconds, receives grid signals from the aggregator's server via a 4G or cloud connection, issues setpoint changes via BACnet/IP. End-to-end response latency: 500 ms–5 seconds depending on polling cycle timing and network latency. Suitable for economic demand response and capacity market participation. Not suitable for frequency-responsive ancillary services.
Architecture B — Edge-first with cloud telemetry: Dedicated edge controller at the building's main electrical panel continuously monitors local voltage and frequency via direct sensor connection or smart meter pulse input. Frequency threshold triggers pre-programmed local control actions — BESS discharge, EV charger curtailment, HVAC stage reduction — via direct digital outputs or sub-100 ms BACnet commands, without waiting for cloud confirmation. Cloud telemetry layer reports event data to aggregator and ISO after the response has already executed. End-to-end response latency: 150–300 ms. Capable of frequency-responsive services including FFR, spinning reserve, and AGC regulation.
Architecture B requires more upfront engineering — the edge controller must be configured with threshold logic, the control sequences must be pre-validated, and the ISO telemetry reporting must meet program-specific requirements. But it unlocks ancillary service revenue streams that Architecture A cannot access, and it provides resilience value (autonomous response during communication outages) that Architecture A explicitly depends on continuous connectivity to provide.
The Economic Case for Fast Response
The premium that ISO markets pay for fast response versus slower response is substantial in markets where the distinction is explicitly priced. In ERCOT, FFR clears at a premium to slower spinning reserve products. In PJM's regulation market, the mileage-adjusted clearing price for Reg-D (fast, dynamic regulation signal following) has historically exceeded Reg-A (slower, traditional regulation) by 30–80% in periods of high regulation demand. These premiums reflect the grid's actual operational need for fast response, not arbitrary market structure.
A 500 kW commercial BESS capable of 200 ms frequency response and enrolled in both FFR and capacity market programs in ERCOT captures measurably higher revenue than the same hardware enrolled only in slower capacity products. The incremental revenue — attributable purely to the fast response architecture — has historically represented 15–25% of total DR program revenue for well-operated distributed storage assets in frequency-sensitive markets.
That incremental revenue is the direct financial expression of a 200 ms response capability versus a 5-second one. The grid needs that speed; it's willing to pay for it; and the engineering to deliver it is within reach of commercial building assets today.