AI Driven Streamlining of Appliance Load Monitoring in Facilities Management

Facilities Management (FM) companies rely on effective and low cost data collection from Appliance Load Monitoring (ALM) devices to provide asset quality and energy monitoring services. The introduction of an automated appliance type classification pipeline during installation and inspection can offer huge improvements in reducing cost and installation errors. Most work focus on showcasing Voltage-Current (V-I) trajectory features based Machine Learning (ML) and Deep Learning (DL) algorithms on benchmarking datasets rather than providing mechanisms for deploying their model onto a production-ready system. This paper introduces a feature extraction preprocessing approach for ensuring the validity of detected steady-state events in VI trajectories that can be used with Machine Learning (ML) models to identify FM asset types during site installations of Appliance Load Monitoring (ALM) units. We introduce a framework in which the approach can be used as part of the training and deployment of ML models for verifying and monitoring assets in FM client environments.