| Summary: | Electromechanical systems experience both gradual and sudden fault conditions. Power monitoring provides a valuable approach for detecting faults, essentially turning a machine into its own sensor for observing developing and abrupt failures. Machines can be monitored individually or nonintrusively (as a collection of loads) and signal processing can tease out relevant indicators of operational status and health. Load identification and diagnostics with aggregate electrical monitoring rely on the correct physical interpretation of measurements. Specifically, ties between the observed measurements and the actual physical task performed by a load ensure the relevance of a measurement, or a feature space derived from the measurement, for reliable identification and diagnostics. This article examines three-phase mathematical relationships for different load configurations, specifically with an eye toward selecting a feature space useful for automated diagnostics. The utility of these three-phase measurement representations is demonstrated with experimental data from several microgrid systems.
|
|---|