Highly Accurate Diagnosis Scheme of Open-Circuit Faults in a Three-Level Inverter Using Optimized Multi-Classifier Decision Fusion

Identifying and isolating open circuit faults in inverters is essential as it greatly enhances reliability, durability, and operational efficiency. This paper presents a data-driven fault diagnosis scheme that combines multiple classifiers to improve the classification accuracy of different open circuit faults in three-level inverters. Statistical features are used to train two supervised machine learning models. Subsequently, a majority voting technique is used to combine the decisions made by these classifiers, selecting the one with the highest prediction probability. The hyperparameters of both classifiers are optimized during the training using Bayesian optimisation. The effectiveness of the proposed method is validated using capacitor voltage signals, eliminating the need for extra sensors. A comparative study is carried out with other conventional supervised machine learning methods, showing that the proposed method is highly accurate in diagnosing various open circuit faults.