8,760 rows · 7 columns

Grid energy load

A year of hourly electricity demand, renewable generation, temperature and wholesale price for one balancing area — including the hours when price goes negative.

What is in it, and what is wrong with it

Load has three nested seasonalities — hour of day, day of week, and time of year — which makes it a much more honest forecasting exercise than the single-seasonality series that show up in tutorials. Prices go negative on windy low-demand nights, so a model that assumes positive prices will look fine on MAPE and be useless in practice.

Schema

ColumnTypeDescription
hour_attimestampHour beginning, UTC.
load_mwfloatSystem demand in megawatts.
temp_cfloatPopulation-weighted temperature.
solar_mwfloatSolar generation.
wind_mwfloatWind generation.
price_eur_mwhfloatDay-ahead clearing price. Can be negative.
is_holidaybooleanPublic holiday flag.

First twelve rows

hour_atload_mwtemp_csolar_mwwind_mwprice_eur_mwhis_holiday
2023-01-01 00:00:004269-2.60219.838.6true
2023-01-01 01:00:004290.6-6.5049337.91true
2023-01-01 02:00:004441.7-60516.620.24true
2023-01-01 03:00:004769.1-5.80603.335.7true
2023-01-01 04:00:005116-3.7051447.96true
2023-01-01 05:00:005938.5-0.10228.945.44true
2023-01-01 06:00:006283.2-1.30103642.15true
2023-01-01 07:00:006808.5-1.9116.6692.555.05true
2023-01-01 08:00:007126.5-3.6154.3764.646.63true
2023-01-01 09:00:007083.8-0.4227.421654.32true
2023-01-01 10:00:007033.62.2401.3372.850.39true
2023-01-01 11:00:006753.86.6467.2373.865.94true