Incremental Classifier Learning for Non-intrusive Appliance Identification

Comprehending the energy consumption of household appliances is essential for making well-informed decisions regarding energy conservation. Non-intrusive load monitoring (NILM) aims to estimate the power usage of individual appliances based on aggregate smart-meter measurements. An ongoing challenge in NILM is the development of learning systems that can incrementally identify new appliances from an online data stream. This study presents an innovative training approach facilitating incremental learning. It requires only a limited library of appliance training data to be concurrently available, allowing for the progressive identification of new appliances. The proposed model's effectiveness has been substantiated through meticulous validation using experimental results from an accessible dataset.