Adaptive on-line unsupervised appliance modeling for autonomous household database construction

A Practical Approach to Residential Appliances on-Line Anomaly Detection: A Case Study of Standard and Smart Refrigerators

“…Careful anomaly detection requires a framework that is capable of continuously monitoring appliances loads and providing their in-operation information for estimation algorithms. Accordingly, durable household load monitoring systems are emphasized as key enabler to designate such a structure [10], [11]. Although, these systems have been thoroughly probed from both intrusive and non-intrusive aspects, their anomaly detection capability has not been fairly taken into consideration.…”

“…In terms of Non-Intrusive Load Monitoring (NILM), few studies have only investigated the proficiency of load disaggregation methods for anomaly detection [12], [13]. Furthermore, in [11], we have aimed to design a NILM system for diagnosis purposes. Nevertheless, state-of-the-art NILM methods are not adequate to provide efficient anomaly detection and thus, diagnosis services [11], [12].…”

“…Furthermore, in [11], we have aimed to design a NILM system for diagnosis purposes. Nevertheless, state-of-the-art NILM methods are not adequate to provide efficient anomaly detection and thus, diagnosis services [11], [12]. In fact, anomaly in electrical appliances has a dynamic stochastic nature, for which providing a training class is a tedious task.…”

“…At the household level, NILM ability to detect anomalies has recently drawn researchers' attention. The authors have previously investigated the NILM capability to provide diagnosis services [11]. Actually, in [11], we have aimed to enable NILM diagnosis capacity by designing a time-variant load modeling system.…”

“…The authors have previously investigated the NILM capability to provide diagnosis services [11]. Actually, in [11], we have aimed to enable NILM diagnosis capacity by designing a time-variant load modeling system. This framework exploits a recurrent pattern recognition and model construction mechanism to capture the dynamic of power consumption.…”