A Hybrid-Model-Flux-Observer-Based Horizon-Adaptive Sliding Integration Fault Diagnosis Method for Sensorless Vector Controlled Voltage Source Inverter

Abstract

In this article, a horizon-adaptive sliding integration fault diagnosis (HSIFD) method based on a hybrid model flux observer is proposed for a single switch open-circuit (OC) fault in the sensorless vector control (SVC) fed voltage source inverter (VSI). It can achieve the diagnosis and location of the OC fault of the power switch quickly in real-time within the frequency regulation range. First, the compensation voltage is obtained in-loop by the deviation between the voltage model flux observer and the current model flux observer. Aiming at the problem that model-based methods rely heavily on precise mathematical models and parameter accuracy, a conversion method from time domain integration to phase angle domain integration is proposed to indirectly observe the periodic characteristics of compensation voltages. Then, the second-order equal phase angle moving mean (SoEPAMM) method is proposed to reduce the data storage and improve diagnosis timeliness. Simulations and experiments show the accuracy and superiority of the HSIFD method in the SVC framework. The proposed method can be implemented in-loop for practical applications since the storage and operation volumes are greatly reduced. Moreover, a comparison with other models shows that the proposed method is robust to the variation of model parameters, speed, and load.