An Explicit-based Model Predictive Control Strategy with Neural Network Fitting for Power Converters

Since the output voltage of a power converter is susceptible to fluctuate caused by load variations, an explicit-based model predictive control (EMPC) strategy with neural network fitting is designed in this paper. The control laws in different operating points are generated offline and stored as tables by the EMPC strategy, and fitted by back propagation neural network. The strategy with neural network effectively not only solves the problem that the traditional MPC strategy requires massive storage space and reduces the computational burden., but also responds rapidly and stabilizes the output voltage while the load fluctuates, improving the robustness of power converters. Finally, the above method is applied in an 14SL-boost converter, and the correctness of theoretical analysis and the feasibility of EMPC with neural network is verified by the simulation.