Neural Network Based Adaptive Model Predictive Control for Power Converters Under Load Parameter Uncertainties

Abstract

This article proposes a new neural network based adaptive model predictive control (named NN-AMPC) for power converters under load parameter uncertainties. Firstly, a supervisor MPC controller is designed for power converter using matched model parameters. Next, a NN is built and trained offline utilizing the operating information from the supervisor controller. A practical adaptive MPC controller using FPGA is then set up utilizing the trained NN to control the power converter online. The proposed NN-AMPC can adaptively track the variation of load parameters without extra identification process of load parameters. The dynamic response of the NN-AMPC under step changes in load parameters are analyzed and compared with conventional MPC. The concept of NN-AMPC is verified by experimental results on a 3-phase voltage source inverter (VSI) as the case study. It is shown that, the FPGA-based NN-AMPC controller offers better dynamic performance in the presence of uncertain parameters while utilizes reduced FPGA resource requirement compared with the observer based MPC controller.