NILM Datasets for Research: IAWE, REDD, and UKDALE

This dataset includes the original IAWE, REDD, and UKDALE datasets, as well as the pre-processed tensor data used in the research paper. The files with the .h5 extension are the original datasets, while the .npy files are the tensorized data used in the paper.

The IAWE dataset is a high-resolution energy consumption dataset collected from a single household in India, recorded at a 1-second interval. It includes data on various appliances such as air conditioners, refrigerators, and lighting systems, making it particularly useful for studying load patterns in developing countries.

The REDD dataset, developed by MIT, is a benchmark dataset in the NILM field. It consists of data from 6 U.S. households, with high-frequency voltage and current waveforms sampled at 15 kHz and low-frequency aggregate power data recorded every second. REDD provides detailed appliance-level energy usage data for devices like refrigerators, microwaves, and washers, making it ideal for energy disaggregation studies.

The UKDALE dataset, provided by Imperial College London, covers long-term electricity consumption data from 5 households in the UK, recorded since 2012. It includes aggregate power data at 1-second intervals and appliance-level data at 6-second intervals, with information on approximately 50 appliances. Its long time span and seasonal variations make it particularly suitable for studying long-term household energy behavior.

The combination of these datasets allows for diverse, real-world energy analysis, while the tensorized .npy files streamline their use in tensor decomposition and deep learning applications.