LPV Model-Based Fault Detection and Isolation in DC Microgrids Through Signature Recognition

The safety of DC microgrids is threatened by multiple types of faults occurring in different components, entailing multi-target fault detection and isolation solutions. To address this issue, this paper introduces a comprehensive fault detection and isolation framework based on the mathematical models of DC microgrids. In this work, the DC microgrids with non-linear characteristics are modeled via polytopic linear parameter varying modeling, which has the remarkable property to adapt to changing operating points. On this basis, a bank of unknown input observers are built. At last, settingless fault classification is achieved through reconstructing the patterns in the outputs of the observers. Compared with existing fault detection and isolation methods for DC microgrids, the proposed system-level solution can simultaneously diagnose multiple faults in DC microgrids in non-stationary operating states. Moreover, the proposed signature recognition scheme solves the difficulty in threshold setting and greatly improves the tolerance of the proposed method to modeling and measurement uncertainties. The accuracy, adaptivity and robustness of this method are verified through extensive numerical tests with MATLAB/Simulink.