Fault Diagnosis and Condition Monitoring of Power Electronic Components Using Spread Spectrum Time Domain Reflectometry (SSTDR) and the Concept of Dynamic Safe Operating Area (SOA)

Abstract

Fault diagnosis and condition monitoring (CM) of power electronic components with a goal of improving system reliability and availability have been one of the major focus areas in the power electronics field in the last decades. Power semiconductor devices such as metal oxide semiconductor field-effect transistor (MOSFET) and insulated-gate bipolar transistor (IGBT) are considered to be the most fragile element of the power electronic systems and their reliability degrades with time due to mechanical and thermo-electrical stresses, which ultimately leads to a complete failure of the overall power conversion systems. Therefore, it is important to know the present state of health (SOH) of the power devices and the remaining useful life (RUL) of a power converter in order to perform preventive scheduled maintenance, which will eventually lead to increased system availability and reduced cost. In conventional practice, device aging and lifetime prediction techniques rely on the estimation of the meantime to failure (MTTF), a value that represents the expected lifespan of a device. MTTF predicts expected lifespan, but cannot adequately predict failures attributed to unusual circumstances or continuous overstress and premature degradation. This inability is due in large part to the fact that it considers the device safe operating area (SOA) or voltage and current ride-through capability to be independent of SOH. However, we experimentally proved that SOA of any semiconductor device goes down with the increased level of aging, and therefore, the probability of occurrence of over-voltage/current situation increases. As a result, the MTTF of the device as well as the overall converter reliability reduces with aging. That said, device degradation can be estimated by accomplishing an accurate online degradation monitoring tool that will determine the dynamic SOA. The correlation between aging and dynamic SOA gives us the useful remaining life of the device or the availability of a circuit. For this monitoring tool, spread spectrum time domain reflectometry (SSTDR) has been proposed and was successfully implemented in live power converters. In SSTDR, a high-frequency sine-modulated pseudo-noise sequence (SMPNS) is sent through the system, and reflections from age-related impedance discontinuities return to the test end where they are analyzed. In the past, SSTDR has been successfully used for device degradation detection in power converters while running at static conditions. However, the rapid variation in impedance throughout the entire live converter circuit caused by the fast-switching operation makes CM more challenging while using SSTDR. The algorithms and techniques developed in this project have overcome this challenge and demonstrated that the SSTDR test data are consistent with the aging of the power devices and do not affect the switching performance of the modulation process even the test signal is applied across the gate-source interface of the power MOSFET. This implies that the SSTDR technique can be integrated with the gate driver module, thereby creating a new platform for an intelligent gate-driver architecture (IGDA) that enables real-time health monitoring of power devices while performing features offered by a commercially available driver. Another application of SSTDR in power electronic systems is the ground fault prediction and detection technique for PV arrays. Protecting PV arrays from ground faults that lead to fire hazards and power loss is imperative to maintaining safe and effective solar power operations. Unlike many standard detection methods, SSTDR does not depend on fault current, therefore, can be implemented for testing ground faults at night or low illumination. However, wide variation in impedance throughout different materials and interconnections makes fault location more challenging than fault detection. This barrier was surmounted by the SSTDR-based fault detection algorithm developed in this project. The proposed algorithm was accounted for any variation in the number of strings, fault resistance, and the number of faults. In addition to its general utility for fault detection, the proposed algorithm can identify the location of multiple faults using only a single measurement point, thereby working as a preventative measure to protect the entire system at a reduced cost. Within the scope of the research work on SSTDR-based fault diagnosis and CM of power electronic components, a cell-level SOH measurement tool has been proposed that utilizes SSTDR to detect the location and aging of individual degraded cells in a large series-parallel connected Li-ion battery pack. This information of cell level SOH along with the respective cell location is critical to calculating the SOH of a battery pack and its remaining useful lifetime since the initial SOH of Li-ion cells varies under different manufacturing processes and operating conditions, causing them to perform inconsistently and thereby affect the performance of the entire battery pack in real-life applications. Unfortunately, today’s BMS considers the SOH of the entire battery pack/cell string as a single SOH and therefore, cannot monitor the SOH at the cell level. A healthy battery string has a specific impedance between the two terminals, and any aged cell in that string will change the impedance value. Since SSTDR can characterize the impedance change in its propagation path along with its location, it can successfully locate the degraded cell in a large battery pack and thereby, can prevent premature failure and catastrophic danger by performing scheduled maintenance.