All production, imbalance, and KÜPST figures are drawn from the EPİAŞ Real-Time Settlement (Gerçek Zamanlı Uzlaştırma / GZU) reports. Data is fetched at hourly resolution via the public EPIAS Transparency Platform REST API (eptr2) and aggregated to monthly plant-level summaries. The dataset covers all licensed solar (GES) or wind (RES) power plants that reported at least one hour of production in the analysis period.

The imbalance ratio measures how much a plant's hourly generation deviates from its day-ahead schedule, relative to what it actually produced. For each hour we take the absolute difference between the forecast (KÜDÜP) and the actual settlement quantity (UEVM), sum these over the month, and divide by total actual generation:Because the hourly differences are taken in absolute value, over-production hours and under-production hours do not cancel out — both count as imbalance. The metric therefore captures hourly forecasting volatility and is scale-independent, so plants of different sizes compare fairly. The separate net imbalance (signed sum, where surplus and deficit hours offset) is typically much smaller and is reported as a distinct figure.

Imbalance cost arises when a plant's actual hourly generation differs from its day-ahead schedule. Settlement is system-direction-dependent: positive imbalance (plant produced more than scheduled) is settled on min(MCP, SMP); negative imbalance (produced less) on max(MCP, SMP).

• Until Dec 2025 — flat 3% margin: positive at min(MCP, SMP) × 0.97, negative at max(MCP, SMP) × 1.03.
• From Jan 2026 (EPDK regulation, Official Gazette 29/12/2025) — the margin depends on whether the plant's deviation aggravates or relieves the system. A deviation in the same direction as the system imbalance is charged 6% (positive × 0.94, negative × 1.06); a deviation that helps the system keeps the 3% margin (positive × 0.97, negative × 1.03).

Figures use EPİAŞ's published hourly imbalance prices, which already embed the applicable system-direction coefficient. The reported monthly value is the total imbalance cost (TL) — every hour is charged on the absolute size of its own deviation, so over- and under-forecast hours add up rather than cancel out.

KÜPST (Kesinleşmiş Üretim Planından Sapma Tutarı — deviation-from-final-schedule charge) is a penalty applied hour by hour, whenever an hour's deviation from the final schedule (KUDÜP) exceeds a regulatory tolerance band set as a share of that hour's schedule:

• Solar (GES): 8% of the hour's KUDÜP (10% before Jan 2026)
• Wind (RES): 15% of the hour's KUDÜP (17% before Jan 2026)

The KÜPST unit price is 5% of max(MCP, SMP, 750 TL/MWh) for the relevant hour from January 2026 (3% until December 2025). Only the portion of an hour's deviation that exceeds the tolerance band attracts the penalty; the monthly figure is the sum over all hours.

Unit cost = (Total Imbalance Cost + Total KÜPST Cost) ÷ Total Generation for the month, expressed in TL/MWh. It provides a scale-independent measure of settlement efficiency, allowing fair comparison across plants of different capacity.

Plants are ordered by their monthly imbalance percentage: the sum of absolute hourly differences between the published forecast (KUDÜP) and settled generation (UEVM), divided by generation, × 100. Hours do not offset each other. The figure reflects forecasting but also curtailment, outages and balancing instructions, which public data cannot separate, so it is not a measure of forecast skill alone. Differences between neighbouring plants are largely statistical noise, so the public table shows a tier rather than an exact rank: a plant is marked lower or higher than the fleet median only when a 95% confidence interval for its imbalance %, estimated by resampling the month's days, lies entirely on that side. Plants whose data does not allow a fair comparison (for example a partial month, very low generation, output far below the unit's usual level, or a day without a published forecast) are not ranked and are listed with the reason.

A plant is excluded from a given month's ranking — while remaining in the registry — in either of these cases:

• Zero activity: the plant reported zero generation and zero forecast for the whole month.
• Insufficient forecast coverage: the plant has both a forecast and an actual value for less than half of the month's hours — typically a unit commissioned or registered mid-month. Its imbalance ratio would not be comparable with a full-month plant, so it is listed rather than ranked.

Every excluded plant is listed below with its reason.

KARAPINAR YEKA-1 (approx. 1,365 MW) is excluded from the plant statistics due to its outsized scale. Its performance is reported separately.

Plants excluded this month

LÜM_İLETİM_GÜLSAN GES_5.1.H(ST)_GÜNEŞ	Data for this unit could not be retrieved or processed for this period — not ranked
ELAZIĞ HİBRİT BES/GES	Zero production and zero forecast for the entire period — not ranked
YENİ GES	Data covers only 672/744 hours of the period — not comparable with a full-period plant, so not ranked
TOZOR GES	Forecast and actual both available for only 79/744 hours — not comparable with a full-period plant, so not ranked
ALİAĞA-1 GES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data
ARTUKLU-1 GES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data
G24-MALATYA GES	The published forecast (KUDÜP) is zero on at least one day with production — the monthly ratio is not comparable with plants forecast every day, so not ranked
IĞDIR DGES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data
ME-SE GES	Imbalance ratio above our 75% review threshold — not assessable from public data, so not ranked
NEPSUN HADİM GES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data
OZE DEPOLAMALI GES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data
PAMUK GES	Not ranked under our policy for storage-integrated and balancing-market units, whose deviations cannot be separated from balancing instructions using public data