Abstract
In this paper, indirect finite control set model predictive control (I-FCS-MPC) is used for detecting, localizing, and tolerating open-circuit failures in the transistors without the use of arm voltage sensors. The fault is detected by the main controller whereas the localization is performed in the local controller which is used for the sorting algorithm. The main controller utilizes the discrete mathematical model to estimate the arm voltages using state measurements from present and previous sampling instants. The arm voltage command given by the main controller in the previous sampling instant is then compared with the estimated arm voltage to detect the fault. The fault signal is sent to the local controller where a counter is increased for the potential faulty submodules (SMs). The fault is then localized to the specific SM whose count first goes above a threshold value. Finally, this SM is bypassed using a bypass switch and a redundant SM is inserted in its place. The proposed fault detection and localization method does not require any additional sensors. Simulation results demonstrate that the fault can be detected, localized, and cleared within one fourth of the fundamental period.
