A Statistical Method Combined with Load Profiling for the Predictive Diagnosis of Anomalies: a Preliminary Analysis
Federico Molinaro, Luca Tari, Domenico Capriglione, Luigi Ferrigno · 2025
The article proposes the use of Smart Monitoring, in particular, Load Profiling, as a measurement method to enable real-time Fault Detection and Identification (FDI) processes of devices and systems. In particular, the paper proposes the optimisation of a Load Profiling method to improve its diagnosis capability through an unsupervised statistical approach and the definition and estimation of new indices and statistical factors designed to describe the presence of anomalies or faults. The main strengths of the proposed work are i) the ability to operate without any physical knowledge of the monitored system and ii) the ability to operate as an early-stage anomaly or fault detector. The key elements of the proposed approach are that it can operate successfully regardless of the complexity and type of the monitored system, can be extended to operate with multiphysical quantities and fields, and can be implemented directly in smart meters.