Fault Detection and Identification in MMCs Based on DSCNNs

Fault detection and location is one of the critical issues in engineering applications modular multilevel converters (MMCs). At present, MMC fault diagnosis based on neural networks can only locate open-circuit a single submodule. To solve this problem, paper proposes localization strategy depthwise separable convolutional (DSC) network. By inputting bridge arm circulating current submodule capacitor voltage into two serially connected networks, not method achieve classification faults, block short-circuit inductance faults MMCs, but it also switch where occur. The simulation experimental results show that proposed achieves locates multiple same arm. This accuracies ≥99% 87.7% for single-point multi-point respectively, which better than some benchmark achievements literature terms accuracy, speed, has fewer model parameters real-time performance.