Wind Farm Curtailment Monitoring

Back to Blog

Why Curtailment Needs Its Own Monitoring View

Curtailment is a deliberate reduction of a wind farm's output below its available power, and it is completely normal. A grid operator may cap export during network congestion, a noise permit may force reduced-power modes at night, shadow-flicker rules may stop specific turbines when the low sun casts a moving shadow on a home, and bat or bird protection programs may stop machines during defined conditions. None of these are faults, yet all of them make the park produce less than the wind allows.

That is exactly why curtailment needs its own monitoring lens. If curtailed periods are not tagged as such, they pollute two important numbers: availability, which should not be penalized for a machine that was healthy but told to stop, and lost production, which should attribute the loss to the right cause. Getting this attribution right is a governance issue as much as a technical one, because contractual availability and warranty claims depend on it. It is the same discipline that makes any state-based reporting defensible in a well-run SCADA, where state categories drive the whole wind farm SCADA system.

The Signals That Reveal and Classify Curtailment

The primary signal is the active-power setpoint the SCADA sends to the turbines, compared against the estimated available power for the current wind. When actual power tracks a setpoint that is below available power, the park is curtailed, and by how much is the difference. Estimating available power is the subtle part: it is inferred from the wind speed and the turbines' own power curves, or from a subset of uncurtailed reference turbines, so the lost-energy figure is always an estimate rather than a measurement.

The classifying signals are the reason codes. A grid dispatch instruction, a noise-mode flag, a shadow-flicker calendar trigger, and a wildlife-protection mode each set a distinct reason on the SCADA so the curtailment can be attributed correctly. A well-configured system stamps the start, end, reason, and estimated energy for every curtailment event, producing an auditable log rather than a vague nightly dip. Because grid-driven curtailment is often anticipated from network and market conditions, it sits alongside the power forecasting the fleet already runs, and it shares the reactive-power and export controls the operator sends down.

Because curtailment estimates lean on inferred available power, practitioners treat them with appropriate humility. The right posture is to report curtailed energy as an estimate, document the method, and reconcile it against the interconnection metering and any grid-operator records. When those cross-checks disagree badly, the available-power model usually needs revisiting, not the meter.

What a Defensible Curtailment Log Contains

Because curtailment accounting feeds availability reporting, warranty positions, and sometimes compensation claims, the event log has to survive hostile review. Each curtailment event should be a structured record, not an annotation someone remembers to add. The fields that matter:

Field	Why it matters
Start and end timestamps	Define the window every downstream calculation uses; time-sync problems here corrupt everything else
Reason code and source	Ties the cap to the grid instruction, permit rule, or wildlife trigger that caused it
Setpoint applied	The actual limitation commanded, which the power trace should visibly track
Estimated available power and method	The counterfactual and how it was computed, so the estimate can be reproduced later
Archived instruction	The dispatch signal or schedule itself, stored verbatim, for disputes

The archived instruction deserves emphasis. When a grid operator's records and the park's records disagree years later, the stored dispatch signal is what settles the argument. The same holds for noise and wildlife modes: the triggering calendar entry or sensor condition should be captured with the event, not reconstructed afterwards.

A Worked Lost-Energy Calculation

The arithmetic is deliberately simple; the judgment lives in the inputs. Take a curtailment window of duration T. For each reporting interval inside it, estimate the power the park could have produced, P-available, from the reference method - power curves applied to measured wind, or uncurtailed reference turbines scaled to the park. Subtract the power actually produced, P-actual, floor the difference at zero, and sum the products of that difference and the interval length across the window. The result is the curtailed-energy estimate for the event.

The floor at zero matters: in gusty conditions the park may briefly produce above the estimate, and letting negative intervals offset positive ones quietly understates the loss. Equally important is versioning the method - if the power-curve set or the reference-turbine selection changes, old events keep the method they were computed with, and the change itself is logged. An estimate that cannot be reproduced is an estimate that will not survive an audit.

Edge Cases That Complicate Attribution

Real parks stack causes. A turbine can fault while the park is under a grid cap, and the accounting must decide whether those hours count against availability or curtailment - the honest answer is that a faulted machine is unavailable regardless of the cap, and the curtailment estimate should exclude it from the counterfactual. Overlapping curtailments need a precedence rule: when a noise mode and a grid cap apply simultaneously, the binding constraint is whichever setpoint is lower, and the log should record both with one marked as binding.

The reference method has its own traps. Reference turbines must themselves be uncurtailed, which fails when the whole park is capped - then power curves and measured wind carry the estimate alone. Wake effects mean a reference machine's scaled output is not a neutral stand-in for every position in the park. None of this is fatal; it simply belongs in the documented method. These attribution questions sit alongside the rest of the park's health picture, from turbine condition monitoring points to the substation and collection-system view in wind farm balance-of-plant monitoring.

Frequently Asked Questions