In contrast to this, the LUKES project proposes aggregate monitoring of 10,000s of homes, while extensively monitoring up to 400 homes for up to 4 years. Although not the primary aim of the study, such a data set would have considerable impact for the energy disaggregation community. I was keen to point out that such uses have arisen due to the Household Energy Study, and it is important to take this data set as a case study when designing new surveys.
However, it is also important that lessons are learned from HES, such that the same mistakes are not repeated. In particular, I’d hope that a new data set would:
- Collect both aggregate-level and circuit-level data as well as appliance level data
- Specify and maintain each household’s metering hierarchy and appliance names using a consistent metadata schema, such as the NILM Metadata project
- Collect aggregate data at a higher resolution than 2 minute energy readings. Ideally, I believe 1 second power data would be best trade-off between cost and frequency