A Model-Data Hybrid Driven Diagnosis Method for Open-Switch Faults in Three-Phase T-Type Grid-Connected Converters

With the widespread application of three-level T-type converters (3LT2Cs), fault diagnosis has become increasingly important. Existing diagnostic methods can be divided into two types: model-driven and data-driven. Model-driven diagnosis is fast and accurate, but defining diagnosis rules can be complicated and difficult, making it less feasible. On the other hand, using artificial neural networks (ANNs) for fault diagnosis is relatively easier, but it requires heavy calculations and a long diagnosis time. To combine the advantages of both methods, this article proposes a model-data hybrid-driven diagnosis method for open-circuit faults in 3LT2C. First, a model is constructed based on the circuit topology. Second, the input and output parameters of the neural network are determined. Finally, the constructed back-propagation neural network (BPNN) is trained using experimental data, based on which a three-layer BP neural network with three inputs and 13 outputs is constructed to achieve open-circuit fault diagnosis. The effectiveness of the proposed fault diagnostic algorithms is verified through experimental results.