Diagnosis of Open-Circuit Faults in Single-Phase Three-Level Rectifiers Under Unknown Parameter Using VDFRLS and Discrete Adaptive Sliding Mode Observer

To address the problem of open-circuit (OC) fault diagnosis in the power switch of single-phase three-level (SPTL) rectifiers under unknown parameter conditions, this article proposes a diagnostic method that integrates the variable direction forgetting recursive least squares (VDFRLS) algorithm with the discrete adaptive sliding mode observer (DASMO). First, a VDFRLS algorithm is proposed, which addresses the problem in traditional least squares (LSs) methods where high correlation between voltage and current data may cause the information matrix to become nearly singular and lead to abnormal parameter updates, thereby obtaining more accurate rectifier’s impedance parameters. Subsequently, a DASMO is designed, which leverages its rapid convergence and reduced chattering to achieve precise system state estimation. An adaptive threshold for fault detection is then developed using the residual current on the grid-side and its norm, ensuring both the accuracy and robustness of the detection algorithm. Furthermore, an innovative fault localization mechanism combining half-wave current sliding window integration with an observer under fault conditions is proposed, enabling precise localization of OC faults in the power switch. Finally, hardware-in-the-loop (HIL) experiments validate that the proposed method is both effective and robust.