Fault diagnosis of modular multilevel converter based on 1DCNN and LSTM network

This paper proposes an open-circuit fault diagnosis method of Modular Multilevel Converter(MMC) sub module based on one-dimensional convolutional neural network(1DCNN) and Long and short-term memory network(LSTM). The front end of the network uses 1DCNN to directly process the original current signal of MMC. Then, these sequences composed of the extracted features are called the input of LSTM. Finally, the MMC simulation system was built, the data was collected for simulation verification, and the comparison and analysis with the other two methods proved that the proposed method not only shortens the running time, but also has a sufficiently high fault diagnosis accuracy.