Blind Non-Intrusive Appliance Load Monitoring

What is it about?

The paper proposes a novel Non-Intrusive appliance Load Monitoring (NILM) method that does not require any training unlike the majority of NILM approaches, supervised or unsupervised, that require training to build appliance models, and are sensitive to appliance changes in the house, and thus requiring regular re-training. Our blind approach uses graph signal processing to perform event detection, data clustering and pattern matching. The experimental results are presented for two datasets (REDD dataset downsampled to 1min resolution and REFIT dataset with 8sec resolution) showing improved performance accuracy compared to an unsupervised Hidden Markov Model (HMM)-based method.

Featured Image

Why is it important?

Disaggregation of a household’s total electricity consumption down to individual appliances using purely analytical tools can deepen energy feedback, support appliance retrofit advice and support home automation. The majority of current NILM approaches, supervised or unsupervised, require training to build appliance models, and are sensitive to appliance changes in the house, thus requiring regular re-training. This paper proposes a method that does not require any training and can start disaggregating as soon as it is introduced in a house. The experimental results obtained using measurements, typical of widely deployed smart meters in terms of granularity and data type, from four houses show high accuracy and its advantage in immediately starting disaggregation.

Perspectives