An Improved Diagnosis Approach for Short-Circuit Fault Diagnosis in MPC-Based Current Source Inverter System

Due to its excellent dynamic performance, the model predictive current control (MPC) method is widely adopted in power inverter systems, such as current source inverters (CSI). However, short-circuit faults in power switches are common in CSI systems, leading to a periodic output current distortion. While previous fault diagnosis methods have primarily focused on CSI systems using conventional control strategies, this study introduces a novel approach based on the MPC method to enhance the reliability of current source converters by diagnosing short-circuit faults in power switches. The proposed method begins by calculating the phase of the reference current vector in real-time. It then leverages the cost function results from the MPC algorithm as the key diagnostic parameter derived through a comparative analysis of predicted system behavior under normal and fault conditions. In the presence of faulty switches, the algorithm identifies the phase of the reference current where the predicted currents align with the reference values. Additionally, it leverages the fact that during a fault, the MPC optimal vector remains unchanged from the moment the fault occurs. The duration and value of this unchanging optimal vector depend on the type and location of the switch fault, enabling precise fault detection and localization using both phase information and the MPC optimal current vector. This approach remains highly robust, as it analyzes both the recovery state, where reference currents are accurately tracked, and the faulty stage, where currents approach zero. To validate the effectiveness of the proposed method, a MATLAB/Simulink model was developed. Simulation results confirm the method’s robustness, as the proposed approach is capable of reliably identifying faults even under low current demand conditions, where conventional methods typically struggle. This is because traditional methods rely heavily on detecting changes in the measured current amplitude, which become increasingly subtle and difficult to distinguish as the current decreases, making fault identification unreliable. This approach provides a reliable and efficient solution for enhancing the safety and performance of CSI systems, particularly in applications where precise fault detection is crucial.