# Would Tarnița-Lăpuștești make money in Romania? A test on real market data

> **Author:** Marius Comper
> **Model:** v1.1.0 — methodology & data-integrity update, September 2026
> **DAM data:** 2019-01-01 → 2026-09-16 (2026 = YTD, 259 days)
> **Canonical URL:** https://mariuscomper.uk/tarnita/en/
> **Romanian edition:** https://mariuscomper.uk/tarnita/

---

## What this model actually tests

A transparent historical-counterfactual and project-economics model, not a feasibility study or an investment-grade bankability model.

**Modelled:** historical day-ahead (PZU) energy arbitrage · storage losses · power and storage constraints · perfect-foresight upper bound · transparent capture and spread sensitivities · simplified capital-recovery economics.

**Not modelled:** full intraday optimisation · balancing/reserve revenue · ancillary-service revenue · endogenous Romanian market re-clearing · transmission constraints · investment-grade financing structure · detailed hydraulic curves · detailed geotechnical risk · project-level EIA.

---

## Headline result (derived from model output)

- **Perfect-foresight upper bound (central RTE case 78.4%), 2019–2025 average: €167.5m/yr** of DAM arbitrage, before capital and operating costs.
- **Simplified capital test (CAPEX €1300m, WACC 8%, 30 years, OPEX €25m/yr): €140.5m/yr.**
- **Break-even capture: 83.9%** of the ex-post optimum — the threshold above which the simplified test clears, before spread compression and unmodelled revenues.
- **2026 YTD (through 2026-09-16): €209.3m** — a partial year, never averaged with full years.
- Mean discharged energy (simulated, not assumed): **2.43 TWh/yr**.

Historical DAM arbitrage alone does not answer whether Tarnița is financeable. The page shows what the energy market would have paid — and what the model leaves unexplained.

---

## Multi-year backtest (central case, RTE 78.4%)

| Year | Mean DAM (€/MWh) | Min | Max | Margin (€m) | Discharged (TWh) | EFC |
|---|---|---|---|---|---|---|
| 2019 | 50.41 | 0.00 | 158.09 | 61.7 | 2.38 | 182.0 |
| 2020 | 39.49 | 0.00 | 150.02 | 46.4 | 2.37 | 181.2 |
| 2021 | 111.42 | 0.02 | 542.50 | 125.5 | 2.28 | 173.8 |
| 2022 | 265.30 | 0.17 | 964.21 | 302.4 | 2.31 | 176.7 |
| 2023 | 103.74 | -23.18 | 436.89 | 169.1 | 2.43 | 185.3 |
| 2024 | 103.51 | -106.36 | 1021.61 | 245.4 | 2.60 | 198.3 |
| 2025 | 108.17 | -100.63 | 603.30 | 222.1 | 2.61 | 199.2 |
| 2026 YTD | 123.06 | -99.75 | 1015.65 | 209.3 | 1.79 | 136.7 |

EFC = equivalent full cycles (discharged energy / 13,100 MWh). Margin = generation revenue − pumping-energy cost, with no credit for leftover stored energy.

---

## Four real days (observed DAM prices, not generated curves)

- 2026-04-26 (15-min): Extreme solar day (26 April 2026) — Actual DAM prices: midday low €-99.8/MWh, evening high €137.7/MWh; spread €237.5/MWh with 35 negative quarter-hours. Midday pumping means being paid to absorb the solar surplus.
- 2024-12-12 (hourly): Winter peak day (12 December 2024) — Actual DAM prices, hourly resolution: peak €843.3/MWh, trough €107.4/MWh, spread €735.9/MWh. A day when every MWh discharged at peak earns its keep many times over.
- 2023-02-11 (hourly): The day the plant should have stood idle (11 February 2023) — Actual DAM prices, hourly resolution: between €129.4/MWh and €161.0/MWh. The peak (€161.0/MWh) never clears the break-even hurdle €165.1/MWh (= €129.4/MWh / 78.4%), so the optimal dispatch is to stand idle: optimal profit €0.
- 2026-03-30 (15-min): Volatile multi-peak day (30 March 2026) — Actual DAM prices: 21 crossings of the daily mean, spread €239.6/MWh. Stop-start several times a day — a test of reversible-unit flexibility.

---

## Key concepts corrected in v1.1.0

1. **Installed power:** 1,000 MW (4×250 MW, historical design basis).
2. **Useful upper volume:** 10 million m³ (historical design).
3. **EPDC five-hour criterion:** documented continuous-operating capability — NOT total storage capacity.
4. **Legacy weekly cycle:** filled and emptied over a weekly pattern, not a simple daily battery.
5. **Derived hydraulic duration:** 10m m³ / 212 m³/s ≈ 13.1 h at maximum documented flow (not an official electrical rating).
6. **Central RTE 78.4%** = 1,649.46/2,103.33 GWh (CNSP 2019/ISPH); 76.2% documented alternative; 72% conservative stress only.

---

## Data and reproduction

- Annual output (JSON): [backtest-annual.json](https://mariuscomper.uk/tarnita/data/backtest-annual.json)
- Representative days (JSON): [representative-days.json](https://mariuscomper.uk/tarnita/data/representative-days.json)
- Assumptions & inspector (JSON): [assumptions.json](https://mariuscomper.uk/tarnita/data/assumptions.json)
- Provenance manifest (JSON): [data-manifest.json](https://mariuscomper.uk/tarnita/data/data-manifest.json)
- Annual table (CSV): [tarnita-opcom-backtest-2019-2026.csv](https://mariuscomper.uk/tarnita/data/tarnita-opcom-backtest-2019-2026.csv)
- Methodology: [Methodology and sources](https://mariuscomper.uk/tarnita/en/methodology/)

The full interval series (92,852 observations) is not redistributed under the data mirror's licence; reproduce it via `scripts/fetch_opcom.mjs` + `scripts/process_opcom.mjs` and validate against the official OPCOM anchors.