python-datons: Query Spain's I90 settlement data with SQL

DATONS_API_KEY=your_key_here

The client reads from either source automatically:

```python
from datons import Client

client = Client()

Explore the data model before writing queries. metadata() returns the schema, available programs, and date range:

meta = client.esios.metadata()

Find units, companies, or technologies with fuzzy search:

client.esios.search("trillo")    # → finds TRL1, C.N. TRILLO
client.esios.search("iberdrola") # → finds company names and units
client.esios.search("solar")     # → finds Solar fotovoltaica, Solar PV...

Three tables are available at different granularities: operational_data_15min (raw 15-min intervals), agg_hourly, and agg_daily (pre-aggregated). Use the coarsest table that fits your query — agg_daily is orders of magnitude faster for monthly or yearly analysis.

Analysis by unit: following one power plant

The per-unit view is where I90 data becomes powerful. One SQL query traces a single power plant through the entire market cascade:

df = client.esios.query("""
    SELECT program, round(sum(energy), 1) AS energy
    FROM operational_data_15min
    WHERE unit = 'BAHIAB'
      AND toStartOfHour(datetime) = toDateTime('2024-12-04 16:00:00')
    GROUP BY program
    ORDER BY program
""")

BAHIAB is a combined cycle gas turbine. At this hour, it scheduled 730 MWh in the day-ahead market — then PHF1 (the first intraday session) added 810 MWh on top. The plant more than doubled its position before delivery. Small corrections from RR and BT, and a net result of 1,526 MWh.

This is gas’s signature move: schedule conservatively, then add volume in intraday. Try the same query with a nuclear unit like TRL1 — you’ll see a flat line. Each technology has a completely different waterfall shape.

Analysis by company: Endesa’s generation portfolio

Zoom out from a single unit to an entire company. What does Spain’s largest generator look like from a technology perspective?

df = client.esios.query("""
    SELECT technology,
           round(sum(total_energy) / 1e3, 0) AS gwh
    FROM agg_daily
    WHERE program = 'PDBF'
      AND day BETWEEN '2024-03-01' AND '2025-02-28'
      AND company_name LIKE '%ENDESA%'
      AND technology IS NOT NULL
    GROUP BY technology
    ORDER BY gwh DESC
""")

The two panels reveal both sides of Endesa’s business. On the right, generation: nuclear dominates, followed by large hydroelectric and combined cycle. On the left, retail: their free-market retailers buy far more energy than their plants produce. Endesa is a net buyer in the day-ahead market — a fact that’s invisible if you only look at generation.

The search() function helps find the exact company name. Company names in I90 data don’t always match trade names — search("endesa") reveals “ENDESA GENERACIÓN, S.A.” as the right filter.

Analysis by technology: the generation mix

The system-wide view shows how Spain’s electricity mix changes through the seasons:

df = client.esios.query("""
    SELECT technology,
           toStartOfMonth(day) AS month,
           round(sum(total_energy) / 1e3, 0) AS energy_gwh
    FROM agg_daily
    WHERE program = 'PDBF'
      AND day BETWEEN '2024-03-01' AND '2025-02-28'
      AND technology IN ('Eólica terrestre', 'Nuclear',
          'Solar fotovoltaica', 'Hidráulica UGH',
          'Ciclo Combinado', 'Gas Natural Cogeneración')
    GROUP BY technology, month
    ORDER BY technology, month
""")

The seasonal rhythm is immediate. Solar peaks in summer and nearly disappears in winter. Wind is strongest in spring and autumn. Nuclear runs flat year-round — the baseload backbone. Hydro fluctuates with rainfall. Combined cycle fills the gaps.

This query runs against agg_daily — the pre-aggregated table. It processes a full year of data across all generation technologies in under a second. The same analysis on raw 15-minute data would scan billions of rows.

Deep dive: who gets curtailed in PDVP?

The real analytical power comes from crossing dimensions. PDVP (technical constraints) is where REE intervenes to maintain grid security. One query reveals who benefits and who pays:

df = client.esios.query("""
    SELECT technology,
           round(sum(total_energy) / 1e3, 1) AS gwh
    FROM agg_daily
    WHERE program = 'PDVP'
      AND day BETWEEN '2024-03-01' AND '2025-02-28'
      AND technology IS NOT NULL
    GROUP BY technology
    ORDER BY gwh ASC
""")

The asymmetry is striking. Combined cycle gas receives massive upward redispatch — REE calls on gas when constrained zones need more power. Every other technology is curtailed. Wind bears the heaviest cuts, followed by solar and hydro.

This is the kind of cross-program, cross-technology analysis that was practically impossible with raw I90 files. Each program was a separate download, each technology had to be filtered manually. With a unified SQL table, it’s one query.

What’s next

Four SQL queries. Four perspectives on Spain’s electricity market that would each take hours of manual data processing. The unified I90 table in python-datons makes this kind of analysis routine.

The library covers I90 settlement data today — every generating unit, every market program, every 15-minute interval since 2023. As the datons platform grows, new datasets will be accessible through the same Client() interface.

pip install datons

---

Full documentation and API key request at datons.com/apps/esios-data/docs.