BLUED: A Fully Labeled Public Dataset for Event-Based Non-Intrusive Load Monitoring Research


The problem estimating the electricity consumption of individual appliances in a building from a limited number of voltage and/or current measurements in the distribution system has received renewed interest from the research community in recent years. In this paper, we present a Building-Level fUlly-labeled dataset for Electricity Disaggregation (BLUED). The dataset consists of voltage and current measurements for a single-family residence in the United States, sampled at 12 kHz for a whole week. Every state transition of each appliance in the home during this time was labeled and time-stamped, providing the necessary ground truth for the evaluation of event-based algorithms. With this dataset, we aim to motivate algorithm development and testing. The paper describes the hardware and software configuration, as well as the dataset’s benefits and limitations. We also present some of our detection results as a preliminary benchmark.

Proceedings of the 2nd KDD Workshop on Data Mining Applications in Sustainability (SustKDD)