The entire US electrical grid contains assets valued at approximately $800 billion, and many of these assets are nearing the end of their design lifetimes. In addition, there is a growing dependence upon power electronics in mission-critical assets (i.e. for drives in power plants and naval ships, wind farms, and within the oil and natural-gas industries). These assets must be monitored. Diagnostic algorithms have been developed to use certain key performance indicators (KPI) to detect incipient failures in electric machines and drives. This work was designed to be operated in real-time on operational machines and drives. For example the technique can detect impending failures in bothmechanical and electrical components of a motor as well as semiconductor switches in power electronic drives. When monitoring power electronic drives, one is typically interested in the failure of power semiconductors and capacitors. To detect incipient faults in IGBTs, for instance, one must be able to track KPIs such as the on-state voltage and gate charge. This is particularly challenging in drives where one must measure voltages on the order of one or two volts in the presence of significant EMI. Sensing techniques have been developed to allowthese signals to be reliably acquired and transmitted to the controller. This dissertation proposes a conservative approach for condition monitoring that uses communications and cloud-based analytics for condition monitoring of power conversion assets. Some of the potential benefits include lifetime extension of assets, improved efficiency and controllability, and reductions in operating costs especially with remotely located equipment.
