Climate tech · Data science · Predict + report

Estimate the sun

Given the weather that happened, how much solar power came onto Great Britain's grid in each half-hour? The grid balances on this number.

Part of Season 1 — Four problems for the planet

upcoming

Cash prizes for the top three. Amounts are announced on Monday 2 November 2026 and paid after the audit. Prizes need a beaten baseline, a filed solution report and a reproducible pipeline — see the rules and terms.

Metric: MAE (lower is better)
Baseline: 700.405; Train-period mean capacity factor for that month and half-hour × installed capacity
Honest model: 279.183; Default gradient-boosted tree on the shipped weather, capacity and the calendar
Closes: 29 Nov 2026, 23:59 UTC; opens 5 Oct 2026, 00:00 UTC
Entrants: 0; 5/day · select 2

Overview

The problem. Great Britain has more than 20 GW of solar panels and almost none of them are metered by the grid. Every half-hour the system operator balances supply against a national estimate of solar output; the better the estimate, the less gas is burned holding reserve. PV_Live, from Sheffield Solar, is that estimate. Your job is the same one: given the weather that happened at twelve points across the country and the installed capacity at the time, estimate national solar generation in megawatts for every half-hour of the test year.

The data. Train: January 2022 to June 2025, 61,296 half-hours. Test: July 2025 to June 2026, 17,520 half-hours. Each row has the timestamp, PV_Live's installed capacity at the time, and NASA POWER hourly weather at twelve points: all-sky and clear-sky irradiance, temperature, cloud amount, wind and humidity. The target is PV_Live's national generation figure. Compute the sun's position from the timestamp yourself.

Scoring. Mean absolute error in megawatts, lower is better. Whole days of the test year are public (30%) or private (70%) so both splits cover every season; your standing is the private split.

Baselines (private split). Predicting the train-period mean capacity factor for that month and half-hour, times the day's capacity: 700 MW. A default gradient-boosted tree on the shipped columns: 279 MW. Beat the first for the badge; beat the second to be in the game.

Why it matters. Every megawatt the estimate misses is a megawatt of reserve someone pays for or a curtailment someone loses. The same skill runs solar forecasting, battery dispatch and grid planning everywhere in the world.

How you are graded

70% the metric, 30% the report. Your selected submissions are scored on the withheld private split at close; your rank there becomes metric points (first of the field = 1, last = 0; below the baseline = 0). Your solution report is graded against the five criteria below by three independent, blind judges — two model families and one variant prompt — at temperature 0; each criterion takes the median level, and any spread of more than one level goes to a human. Final points = 70 × metric points + 30 × report score. Medals and prizes need a beaten baseline and a filed report; the top ten are audited by re-running their code from train.csv.

Criterion	Level 2 — Solid	Level 3 — Exemplary
Problem framing and impact	States the decision, the user and what a given error costs them.	As 2, and quantifies the value of the improvement over the baseline in the user's terms.
Data understanding and validation	A split that mirrors the test split; leakage risks named and handled.	As 2, with evidence the local validation tracks the public score, and a data finding that changed the approach.
Method and rigour	Choices justified, compared against both pinned baselines, at least one thing that failed reported honestly.	As 2, with an ablation or sensitivity check that shows which parts matter.
Error analysis and limitations	Shows where the model fails (segments, conditions) and gives a plausible reason.	As 2, and states the deployment risks: who is harmed if it is over-trusted, what would have to be monitored.
Reproducibility and clarity	Code, environment and steps that a reader could run; prose is clear and within length.	As 2, with fixed seeds, a one-command run, and results that match the report.

All four levels, the judge prompts, model ids and the calibration statistics are published with the results. Rules and terms.

Timeline

Mon, 5 Oct 2026OpensData released; submissions accepted.
Mon, 2 Nov 2026Mid-pointPrize pool announced; first clinic.
Sun, 29 Nov 2026ClosesReport and selections frozen; final 48 hours are the quiet period.
Mon, 7 Dec 2026ResultsPrivate standings, medals and the transparency pack. Seven days to appeal.

Data

Sources and licences

PV_Live, Sheffield Solar — GB national half-hourly PV outturn (pes 0) with installed capacity — Creative Commons Attribution 4.0 International (CC BY 4.0) · retrieved 2026-09-21
PV_Live by Sheffield Solar, University of Sheffield — CC BY 4.0. https://www.solar.sheffield.ac.uk/pvlive/
NASA POWER hourly meteorology and solar irradiance at twelve GB points (MERRA-2 / CERES SYN1deg) — NASA Earth-science data: full and open sharing, no restrictions on use (US government work) · retrieved 2026-09-21
Data obtained from the NASA Langley Research Center POWER Project, funded through the NASA Earth Science Directorate Applied Science Program.

Split. January 2022 to June 2025 is train; July 2025 to June 2026 is test. Within the test year, whole days are public (30%) or private (70%) by a hash of the date, so both splits cover every season and no half-hour has a neighbour in the other split.

Changes made. PV_Live half-hours joined to the NASA POWER hour that contains them (both half-hours of an hour carry that hour's weather); NASA missing values (-999) blanked; rows shuffled and given opaque ids; the target removed from the test file.

Withheld. Nothing beyond the target for the test year.

Rows: train 61,296 · test 17,520 (public 5,280, private 12,240).

Artefact audit passed: No duplicate rows across train and test (0 of 17520 test rows have an identical feature row in train); 1-NN does not beat the honest model (1-NN mae 483.8332 vs honest 282.4693); No single feature encodes the target (strongest |corr| 0.954 (birmingham_ghi)); Row order is uncorrelated with the target (spearman(row index, target) = 0.006); The id does not encode the target (spearman(id number, target) = 0.004).

Half-hourly weather does not exist in NASA POWER; the hourly value is repeated for both half-hours. The datetime is the feature the sun's position hangs on — compute solar elevation from it.

id	Row id (one per half-hour). Use it in your submission.
datetime_utc	Start of the half-hour, UTC. Derive hour, day of year, sun position and daylight-saving from it yourself.
installed_capacity_mwp	GB installed PV capacity in MWp at that time (PV_Live's estimate). It grows through the window.
<point>_ghi	All-sky surface shortwave downward irradiance for that hour at the point, Wh/m² (NASA POWER). Twelve points: london, birmingham, manchester, leeds, bristol, norwich, southampton, plymouth, cardiff, newcastle, glasgow, edinburgh.
<point>_clearsky	Clear-sky irradiance for the same hour, Wh/m² — what the sun would give with no cloud.
<point>_temp_c	Air temperature at 2 m, °C.
<point>_cloud_pct	Cloud amount, %.
<point>_wind_ms	Wind speed at 10 m, m/s.
<point>_rh_pct	Relative humidity at 2 m, %.
generation_mw	Target. GB solar generation in MW for the half-hour, PV_Live's national estimate (train only).

Rules

One entry per person; your handle is fixed once you join. Entering means you accept the competition rules and terms.
Up to 5 submissions a day (resets 00:00 UTC). Select up to 2 for final scoring; if you select none, your best public score's file is used.
Train on the shipped data only. No external data beyond what the data card lists, and never the withheld targets from any source — looking them up is disqualification.
Your solution report (30% of your points) is written on this page and graded as it stands at close. Medals and prizes need a filed report and a beaten baseline; the top ten are audited by re-running their code from train.csv.
Sharing code or ideas publicly is fine; sharing a submission file is not (identical files are detected). One account per person. Text aimed at the grader scores zero.
The final 48 hours are the quiet period: the public board freezes for viewers while you still see your own scores. Results are published once, after the audit, and are final subject to appeal.

The full competition terms apply to every entry.