Condition Monitoring of MMC Submodule Semiconductors

Abstract

Submodules are the building blocks for MMC-type HVDC converters and therefore of utmost importance for converter reliability. In order to unlock smarter maintenance strategies for submodule semiconductors, the technical condition of the semiconductors needs to be estimated to form a health index. Besides the non-project-specific traditional health index methods, a complete practical implementable data-driven lifetime estimation approach for individual converters is needed to determine the exact remaining useful lifetime of all submodule semiconductors. The methodologies should finally explore a new research direction in the optimization of the converter lifetime considering maintenance intervals. This thesis presents a health index methodology applicable to submodule semiconductors in modern MMC-type HVDC converters. The ON-state collector-emitter voltage is used as a condition indicator and external influences are neglected with electrical and thermal models. These models are needed to estimate the junction temperature in the submodule semiconductors due to the temperature-dependent ON-state collector-emitter voltage. A novel lifetime optimization methodology is presented that determines the submodule that is near its end-of-lifetime and should be replaced in the next maintenance interval. In order to optimize the converter lifetime, the selected submodule is switched more frequently based on a selection window generated from the submodule insertion vector. Switching this submodule more frequently can ensure that it fails exactly at a determined time instant and allows other submodules to be switched less frequently extending their remaining useful lifetime. Finally, a complete data-driven lifetime estimation methodology has been described that uses available measurements in existing MMC-type HVDC converters. Based on the converter load profile and the semiconductor specifications, estimations are made on the submodule semiconductors’ lifetime. Simulations show the significant influence of the DC-pole current based on different voltage class semiconductors and their limitations considering the minimum required lifetime.