A battery needs a real optimiser because state of charge couples every interval. A wind or solar farm has no state variable — each interval is an independent decision: curtail whenever the expected price is below your threshold T. T is your marginal opportunity cost: a purely merchant farm uses $0. A farm selling LGCs at spot keeps generating down to about −(LGC price), since every curtailed MWh also forfeits a certificate — but LGC spot has collapsed to roughly $5/MWh (mid-2026, from ~$45 historically), so that case is now thin; fixed-price bundled PPAs can imply much deeper thresholds. Pick T with the control below — there is no single right answer, so we don't pretend there is.
Forward view: AEMO PREDISPATCH 30-min regional prices, refreshed each rebuild. Per-farm UIGF forecasts aren't in the PredispatchIS report, so the farm's expected output is its own recent average daily shape (latest month with at least 20 days of SCADA — never the current partial month's few days) — an availability proxy, not a weather forecast. Numbers are planning-scale indications, not dispatch instructions.
Scorecard: perfect-hindsight bound — revenue the farm would have kept had it curtailed every interval below T, versus what its metered output actually earned. Observed SCADA already includes whatever curtailment (economic or network) the operator did, so the uplift shown is the residual left on the table. Semi-scheduled units implement curtailment through their bids/rebids, not an off switch. Spot-only: no LGC, FCAS, PPA or hedge cashflows.
Sorted by uplift — revenue kept by perfectly curtailing below T. Spot-only, hindsight bound.