Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of GraphSearch.
Among the policies to reduce the consequences of GHG emissions, the decarbonization of the electrical, transport, and manufacturing sectors has profoundly changed how the electricity is produced, transmitted, distributed, and consumed, particularly imposing several complex technical challenges to power electronics-based technologies. Future power converter solutions are called to fulfill increasingly demanding requirements of efficiency, power quality, cost, volume, and reliability, foreseeing that the next generation of high-power converters might be based on modern concepts such as the MMC.Although the MMC exhibits extraordinary characteristics, namely, full modularity, voltage-current scalability, and high efficiency, its development is still in an early stage of maturity, and different topics such as reliability improvement are still research subjects. Around 40000 parts composing the 96 SMs used in the two MV 250kVA MMC to form the PEL MMC research platform illustrates the reasonable concern about technology reliability. CHM arises as an attractive concept to prevent failure events improving reliability and availability of power electronics-based converters. When intended for MMC, CHM development and deployment face several challenges, mainly due to the existing technique's demand for additional hardware, modification of control schemes, particular operating conditions, and complex algorithms. Consequently, MMC CHM state-of-the-art is narrowed to SM power capacitors and power semiconductors. This thesis aims to provoke and contribute to the CHM body of knowledge by investigating three alternatives applied to the PEL MV MMC. The first is a simple, accurate, and costless online strategy to monitor the SM electrolytic capacitor's condition. The method basis on the relationship between the capacitor's degradation and capacitance level. Thus, the SM capacitance is estimated by employing measurements commonly used for necessary control and protection algorithms and the RWLS technique. RT-HIL simulations and experimental results show the method's performance under different capacitor aging levels and converter operating conditions.A second alternative responds to the difficulties in gathering information about components other than power capacitors and power semiconductors, as cost, space, and data processing restrictions are common barriers. The strategy uses the single switch flyback-based ASPS consumption monitoring to observe simultaneously different SM sub-circuits instead of a particular component and the fault dictionary concept to detect the SM drifting from expected healthy conditions. It is demonstrated that minor and significant power consumption variations can be noticed and used for health monitoring.Envisioned in future MMC, the handling, processing and extraction of valuable information from massive amounts of data coming from different CHM techniques is an open issue and certainly a challenge for the health assessment. Inspired by the data fusion framework and MCDM problems, a scheme to integrate various SM health indicators is proposed to compose a comprehensive health index. Systematic approaches considering objective data from the SM and subjective information from expert knowledge are presented and verified through numerical examples based on experimental data. It is demonstrated that entropy, fuzzy-TOPSIS, and game theory-based methods are superior solutions for the SM and converter-level health assessment.