Please use this identifier to cite or link to this item:
|Title:||Reliability analysis and improvement of multilevel converters||Authors:||Tu, Pengfei||Keywords:||DRNTU::Engineering::Electrical and electronic engineering||Issue Date:||3-Jun-2019||Source:||Tu, P. (2019). Reliability analysis and improvement of multilevel converters. Doctoral thesis, Nanyang Technological University, Singapore.||Abstract:||Power converter based variable-frequency drive is much more energy efficient than traditional fixed-speed motor. It can reduce the operating cost of motor drive system significantly. However, the reliability of power converter is a salient concern for both manufacturers and end users as high failure rate will incur additional repairing cost. In high power drive applications, multilevel converters which utilize mature power semiconductors are superior to conventional two-level converters in efficiency and power quality. But the reliability problem of multilevel converters is more serious due to a large number of vulnerable power semiconductors used. Therefore, reliability analysis and improvement of multilevel converters are indispensable for popularization and application of variable-frequency drive. Traditionally, there are two branches in reliability modeling: failure rate methods and Physics-of-Failure methods. The data-driven statistical failure rate methods are widely used in power system for their simplicity and effectiveness. There is a crucial assumption that the failure rate is constant during equipment useful lifetime. The assumption is reasonable in power system where equipment is operating in a relatively stable condition and under good maintenances. However, due to harsh environment and variable load in drive systems, the equipment stress change severely and rapidly leading to a questionable assumption of constant failure. A load-dependent failure rate method is proposed to model the variable load. In the proposed method, the failure rate can be represented as an equivalent constant value or a time-varying function depends on the application. In traditional failure rate methods, the reliability index, Mean Time to Failure (MTTF), is widely used to give a life expectancy. MTTF is the reciprocal of system failure rate in the condition of constant failure rate. With time-varying failure rate, MTTF is almost impossible to calculate in practice. A new reliability index, MTTF consumption is proposed to describe the state of health of a power converter. Multilevel converters with modular design have inherent redundancy which gives them fault-tolerance capability. The network reliability modeling techniques, combinatorics and stochastic process, are used to model the inherent redundancy in multilevel converters. With the aforementioned techniques, the complete and systematic load-dependent failure rate method can be applied to analyze multilevel converter reliability. The Monte Carlo simulation is used to verify the results of load-dependent failure rate method. On the other hand, Physics-of-Failure methods which model failure mechanisms draw lots of attention in the power electronics community recently. These methods calculate components lifetime consumption under a mission profile with pre-established life-stress models. The converter system lifetime is determined by the weakest component. The Physics-of-Failure methods give a deterministic lifetime with clear physics-based models. However, multilevel converter redundancy and system-level reliability are hard to model as there is no component lifetime distribution. The randomness of failure and probabilistic property of reliability are lost. A probabilistic Physics-of-Failure method is developed by including uncertainty of component parameters and life-stress model parameters. An assumption that parameters follow normal distributions is used in the proposed method to simulate the manufacturing process. Component lifetime distribution can be obtained by repetitive Physics-of-Failure analyses or a Monte Carlo simulation. The component reliability function can be extracted by probability distribution fitting. The multilevel converter redundancy and system-level reliability can be modeled by well-established statistical techniques. This thesis provides reliability analysis methods for power converters and especially, multilevel converters and explores possible measures to improve reliability and reduce cost. The thesis has three parts. Part 1 has two chapters with chapter 1 introducing the multilevel converters and its reliability problems, and chapter 2 reviewing reliability modeling history and methods. The second part discusses the two proposed reliability modeling methods with case studies. Improving reliability in component level by switching loss optimization is also presented in this part. The last part draws the conclusions and proposes some future works.||URI:||https://hdl.handle.net/10356/90272
|Appears in Collections:||EEE Theses|
Files in This Item:
|Reliability Analysis and Improvement of Multilevel Converters.pdf||PhD Thesis||7.06 MB||Adobe PDF|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.