Supervisor:
Professor Frede Blaabjerg

Co-Supervisor:
Assistant Professor Saeed Peyghami

Assessment Committee:
Associate Professor Amin Hajizadeh, AAU Energy (Chair)
Professor Amirnaser Yazdani, Ryerson University, Canada
Professor Dirk Van Hertem, KU Leuven, Belgium

Moderator:
Professor Francesco Iannuzzo

Abstract:

The main function of the electrical power system is to deliver the demanded power to the end customers in a reliable and economic way. As a result, the reliability is an important requirement in all power systems, since any interruption in the supply of electricity can have large socio-economic consequences. At the same time, the power systems are undergoing a substantial transition, by gradually retiring the conventional synchronous generators and integrating many renewable generation and storage units. Notably, all these units have power electronic converters as a key element, which enables processing power and grid integration. Therefore, the modern power systems will be Power Electronic-based Power Systems (PEPS). On the other hand, there are some reports of noticeable failure rate of power converters in the field. Further, these failure rates increase over time, due to wear-out and aging of components, which aggravates the situation, as predicted in reports by some Transmission System Operators (TSOs). Thus, the large integration of power converters into the grid have raised concerns in terms of the log-term system-level reliability of PEPSs.

To address these concerns, the first step is to be able to assess the PEPS reliability quantitatively. In other words, a method must be developed to enable calculating the system-level reliability of PEPSs. Nevertheless, to achieve this, there are research gaps and challenges that must be investigated and addressed. In this regard, conventional power system reliability assessment methodologies cannot be used directly for PEPSs, due to their oversimplifications and inherent mathematical limitations. For example, they use a purely statistical approach to extract a constant failure rate for modeling the synchronous generator outages, which neglects the aging of units. On the other hand, outage of power electronic converters is a function of their mission profiles, lifetime models, control strategy, and design parameters. The power converter outages can be modeled by considering these factors together with the physics of failure, which results in non-constant failure rates that reflect the power converters aging. However, well-known mathematical methods used for power system reliability assessment (such as Markov method) are unable to be used with these considerations. Therefore, new mathematical methods must be developed to comply with the new consideration introduced in PEPS. Also, various sources of uncertainties in the system, including, generation, mission profiles, and component-to-component variations, must be considered and modeled.

Hence, in this PhD thesis, a comprehensive framework has been proposed to address the above challenges and enable calculating the stem-level reliability of PEPSs with realistic considerations. Further, the thesis gives a good understanding of “power electronic” and “power system” reliability modeling and assessment fields, by highlighting their differences, similarities, assumptions, and limitations. Subsequently, guidelines are proposed to merge the gaps between these field and achieve the system-level framework for PEPSs. The proposed framework is consisting of several blocks, including availability modeling, scenario generation, power system modeling, state enumeration, and index calculation, where the functions and details of each block have been explained. Moreover, the system-level reliability of several case study PEPSs, have been assessed and analyzed to provide insights and demonstrate the capabilities of the proposed framework.

Notably, the outcome of the PhD project is a model-based framework developed based on a V-shaped approach, where the effect of parameters from the component-level up to the converter-level are reflected on the system-level indices, and contrariwise. Furthermore, the models that are based on physics of failure are mission profile-dependent. Also, the developed methods are computationally efficient, which is critical for analyzing larger system with more power electronic units. Moreover, the framework is developed according to a hybrid “time-based” and “probability-based” approach, by presenting time-dependent PDFs (Probability Density Functions) for reliability indices. The time-dependent term enables muti-timescale analysis of the reliability indices – e.g., investigating the impact of converter aging on the yearly reliability index, or the impact of generation uncertainty on the monthly variation of the reliability indices. The PDFs also represent various sources of uncertainties that exist in the system. Also, the maintenance and change failures are considered in the developed methodologies.

Therefore, the proposed framework provides a tool for system operators to evaluate their system reliability quantitatively. By doing so, they can benchmark the system and ensure whether it meets their long-term goals or if corrective measures must be taken. Also, it will be helpful to system designers, as it enables not only assessing the current status of the system, but also predicting the future performance of the system. Thus, different design scenarios can be evaluated and benchmarked in terms of long-term reliability. Furthermore, the Design for Reliability approach can be realized at the system-level, since a quantitative methodology exists, which enable calculating the proper design margins. In this regards, the system-level reliability indices of several case studies are calculated and analyzed by using the developed method. The impact of power converter aging on the system availability and outage duration and severity are quantified and analyzed. Furthermore, the influence of uncertainty and temporal patterns of mission profiles on the system-level reliability are investigated by providing time-dependent PDFs for the indices. By analyzing the system-level reliability indices, it was shown how the method can be used for benchmarking the system, design, and ensuring the long-term reliability of PEPSs.