A Neutral‐Point‐Voltage‐Residual‐Driven Multi‐Hypothesis Diagnosis Method for Open‐Circuit Faults in Three‐Level NPC Inverters

Accurate and rapid localization of single‐device open‐circuit faults remains challenging in three‐level neutral‐point‐clamped (3L‐NPC) inverters. This paper proposes a neutral‐point‐voltage residual‐driven multi‐hypothesis diagnosis method. The method detects open‐circuit faults by exploiting the imbalance and drift of the dc‐link neutral‐point voltage induced by the fault, and it narrows the candidate set to six switches using the drift polarity index. For these candidates, open‐circuit fault hypotheses are formulated to generate hypothesis‐specific predictions of the neutral‐point voltage, from which residuals are computed. The faulty switch is pinpointed by selecting the hypothesis with the minimum residual, yielding the optimal hypothesis index. After fault detection, a fault‐triggered perturbation is introduced into SPWM to enforce a prescribed diagnostic vector sequence, thereby improving residual separability. This mechanism intentionally constrains the switching sequence so that the operating vectors are steered to an optimal diagnostic vector set, thereby enhancing inter‐hypothesis discriminability and reducing localization time. Experimental results demonstrate that the proposed method requires no additional sensors or hardware modifications, incurs low computational burden, and achieves fast detection and precise localization. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.