Sensorless Model Predictive Control Using Artificial Neural Network for Multilevel Flying Capacitor Boost Converters

In the article, one new sensorless model predictive control method using artificial neural network (ANN-SMPC) is presented for multilevel flying capacitor boost converter (FCBC) to address the issue of over-reliance on flying capacitor (FC) voltage sensors. Firstly, the sampling data obtained from simulation environment is used to train the ANN offline, then the trained ANN is applied to the MPC controller instead of the FC voltage sensors for multilevel FCBC. The ANN structure, data selection and training method of ANN-SMPC are introduced in detail. Its feasibility is proved by simulation and test results. The FPGA-based ANN-SMPC controller can provide control performance comparable to traditional MPC while significantly reduce the FC voltage sensors.