Web app

openenergy.app is a small FastAPI app on the Python API. It values a battery on a historic GB year or on a NESO FES pathway's modelled prices, answers in one plain sentence, and shows the charts, assumptions and sources behind it. It runs at open-energy-eight.vercel.app, or locally:

uv sync                        # the app dependency group (FastAPI, plotly, uvicorn) is a default
uv run uvicorn openenergy.app:app

Then open http://127.0.0.1:8000/. The app reads the repository's data/ folder, so run it from a clone; pip install openenergy[app] installs the code and its dependencies only.

Page	What it does
/	Form: historic GB year (2017–2025, Ember prices) or a FES pathway in 2030 or 2035; power (MW) and duration (0.5–8 hours); forecast (last week's prices or perfect foresight); frequency response (none, Dynamic Containment, or Containment and Moderation; historic years from 2022)
/result	The answer: one sentence (revenue for your MW, share of perfect foresight, cycles a day), key numbers, cumulative and monthly revenue charts, what the result assumes, sources, and links to download it. At once for 1, 2 and 4 hour batteries; otherwise a progress counter while it runs
/download/daily.csv	Daily revenue and perfect-foresight revenue for your MW, with the data credit on the first line
/download/scenario.yaml	The scenario that reproduces the result with openenergy run (or openenergy system run for a FES pathway) from the repository folder
/explore	Battery value by GB year (2017–2025) and by FES pathway (2030, 2035) for one duration, forecast and response option, as bar charts and tables; click a bar to open its result
/compare	Up to three results side by side (table, revenue through the year, and each answer); the page address holds the comparison, so it can be shared, e.g. /compare?item=gb-2019&item=gb-2022&item=falling_behind_2030
/battery	The full hourly report of openenergy report (runs the backtest now)
/scenarios	Each FES pathway's modelled results on 2025's weather next to NESO's own figures
/api/result	JSON: per-MW summary and daily revenue for the same request (without power_mw); see below
/health	Status and version

Every input and key number has a "?" next to it that explains it in a sentence (hover, tab to it, or tap). The battery uses 95% efficiency each way and a £5/MWh wear cost. On Vercel a run took 18 seconds for GB 2024, 14 seconds for a FES pathway and 44 seconds with Dynamic Containment (measured 14 October 2026). Pathway prices are precomputed in data/scenarios/ by scripts/build_scenario_prices.py (see Forward scenarios), so the app does not need PyPSA. Inputs are bounded query parameters; the app reads only the bundled data, from $OPENENERGY_DATA or ./data.

JSON API

/api/result takes the same parameters as /battery (source, year, pathway, target, hours, forecast, services) and returns results per MW, because revenue scales with power at a fixed duration:

{
  "key": "gb-2024-2h-naive_last_week-none",
  "label": "2 h battery, GB 2024",
  "precomputed": true,
  "summary": {"days": 366, "revenue_per_mw_year": 29039.2, "perfect_per_mw_year": 39304.0,
              "capture_ratio": 0.739, "cycles_per_day": 1.57, "captured_spread": 26.72,
              "forecast_mae": 19.54, "response_share": 0.0},
  "daily": {"date": ["2024-01-01", "..."], "revenue": [111.27, "..."],
            "benchmark_revenue": [241.64, "..."]}
}

1, 2 and 4 hour batteries on every year, forecast, response option and FES pathway are precomputed in data/app/ by scripts/build_app_results.py and return at once; other durations run live. A request that cannot run (for example frequency response before 2022) returns status 400 with an error message.

Vercel

The repository deploys to Vercel as one Python function. app.py at the root imports openenergy.app:app; Vercel installs the project with uv, including the app dependency group (default groups other than dev); vercel.json gives the function up to 120 seconds and leaves out tests, docs, scripts and data files the app does not read.