PV-Augmented NILM Dataset (Hugging Face Dataset Card Release)

Directory layout
- REDD_house1, REDD_house2, REDD_house3
- UKDALE_house1, UKDALE_house2

Files per house
- train.csv, test.csv, full.csv
- train.pkl, test.pkl, full.pkl
- train.h5, test.h5, full.h5 (NILMTK-compatible)

Data formats
- CSV: first column is the UTC timestamp index (e.g., 2011-04-19 00:00:00+00:00).
- PKL: pickled pandas DataFrame with the same columns and index as the CSV files.
- H5: NILMTK-compatible dataset built from the full set. Mains are stored as
  aggregate_power_with_injection (active) and aggregate_reactive (reactive).
  Appliances are stored per-column; micro_inverter is exported as device model
  solar_pv. The same export logic is applied to train.h5 and test.h5.
  Columns ending with _state or the helper columns below are excluded from H5.

Common columns (all houses, CSV/PKL)
- aggregate_power
- aggregate_power_with_injection
- aggregate_reactive
- hour_sin
- hour_cos
- power_factor
- micro_inverter
- micro_inv (PV on/off indicator)
- micro_inverter_normalized
- GHI
- DNI
- DHI
- Wind Speed
- Temperature

Appliance columns by dataset (CSV/PKL)
- REDD_house1 / REDD_house3
  - microwave, microwave_state, microwave_reactive
  - fridge, fridge_state, fridge_reactive
  - dish washer, dish washer_state, dish washer_reactive
  - washing machine, washing machine_state, washing machine_reactive
- REDD_house2
  - microwave, microwave_state, microwave_reactive
  - fridge, fridge_state, fridge_reactive
  - dish washer, dish washer_state, dish washer_reactive
- UKDALE_house1 / UKDALE_house2
  - kettle, kettle_state, kettle_reactive
  - microwave, microwave_state, microwave_reactive
  - fridge, fridge_state, fridge_reactive
  - dish washer, dish washer_state, dish washer_reactive
  - washing machine, washing machine_state, washing machine_reactive

Usage examples
- CSV
  - import pandas as pd
  - df = pd.read_csv('full.csv', index_col=0, parse_dates=True)
- PKL
  - import pandas as pd
  - df = pd.read_pickle('full.pkl')
- H5 (NILMTK)
  - from nilmtk import DataSet
  - dataset = DataSet('full.h5')
  - mains = dataset.buildings[1].elec.mains()
  - meters = dataset.buildings[1].elec.submeters()