Room-Level Electricity Consumption Dataset of a Bangladeshi Household

This dataset contains residential electricity consumption and indoor environmental data collected from a single-family unit in an apartment complex. A Refoss EM06P 6-channel smart energy meter captures electrical measurements across six circuits: main grid supply, backup power (Diesel Generator), and four individual rooms. Each channel reports current, voltage, active power, power factor, and cumulative energy consumption. The EM06P publishes these readings over a local MQTT connection to a Raspberry Pi 4 every 5 seconds.

A temperature and humidity sensor connected directly to the Raspberry Pi 4 provides indoor environmental context, sampling at the same 5-second rate and enabling correlation between ambient conditions and electrical load. The temperature sensor measures the temperature at the center of the house, where active cooling, such as airflow from ceiling fans, is typically less used. Thus, the measured temperature may be higher than the temperature experienced in occupied zones with active air circulation. Nevertheless, these measurements provide a useful indication of the baseline indoor thermal conditions. They can help estimate the cooling demand required to maintain an acceptable level of thermal comfort for occupants.

Data structure: Data are stored as one CSV file per day. All timestamps are in UTC (ISO 8601, e.g., 2026-08-31T18:00:02.474Z); Bangladesh time is UTC+6.

• Power logs (DD-MM-YYYY.csv): 31 columns, namely Timestamp and, for each of the six channels (column suffix 1 to 6), V (RMS voltage, V), I (RMS current, A), P (active power, W), PF (power factor, unitless, 0 to 1) and DayE (energy accumulated since the start of the day, kWh).
• Temperature and humidity logs (aht20-DD-MM-YYYY.csv): Timestamp, Temperature_C (air temperature, °C) and Humidity (relative humidity, %RH), measured with an AHT20 sensor.

Raw daily files may contain missing samples (for example during load shedding or logger downtime), repeated header rows, or duplicate records. Helper scripts in the repository merge the daily files, remove duplicates and invalid rows, and optionally insert rows for missing timestamps on a regular 5-second grid (power values set to zero, temperature and humidity left blank), flagged by a Filled column.

Link to dataset: https://github.com/sayeed97h/CDATA-BDHS

Data available since mid-August 2026; repository updates with new data every 3 months. Cleanup and formatting/merging scripts are available in the repository. Scripts generated using generative AI (Claude Sonnet 5.5, Anthropic).

Acknowledgment: Remote data retrieval from the logger uses IT resources from the Power Electronics Group of TalTech (relevant funding information below)

Funding Agency and grant number:

PRG3054: "Direct Current Electrification of Buildings: From Concept to Implementation", funded by the Estonian Research Council