Adaptive Threshold-Based Local Fault Diagnosis Method for DC Microgrids

Line short-circuit faults diagnosis in dc microgrids (dcMGs) requires high rapidity and reliability due to their low inertia and weak overcurrent capability. At present, threshold-based diagnosis methods are popular due to their fast diagnosis speed and ease of employment. However, the existing diagnosis thresholds are usually fixed and determined by simulation and empirical analysis, which lack flexibility and adaptability. In addition, they usually depend on additional line current and/or voltage sensors, which may fail due to sensor faults. To solve the above issues, an adaptive threshold-based fault diagnosis method using information shared with the local converter controller is proposed. First, the difficulties of diagnosis threshold setting under fewer sensors are revealed. Second, the skewness, cumulative sum, and multistep derivatives of single-end converter bus-side current and voltage are selected as the fault features. Adaptive detection thresholds are designed to improve the accuracy and rapidity of line fault identification without line current/voltage sensors and communication links. Third, a modified fault classification method based on real-time calculated quantile interval and bus-side voltage estimation is presented. The accuracy and flexibility of line fault classification can be enhanced. Finally, the performance of the proposed method is validated through MATLAB/Simulink simulations and experimental tests.