Neural network based controller design for three-phase PWM AC/DC voltage source converters

Three-phase AC/DC converter is widely used in many industrial applications. To improve performance, this paper proposes an adaptive neural network based controller design for three-phase PWM AC/DC voltage source converters. The controller is designed based on a nonlinear multi-input multi-output model using Lyapunovpsilas direct method. Since neural networks can approximate unknown nonlinear dynamics, there is no need to know the parameters of the system. In this way, the controller is robust to parameter drifting and changes of operating points. Additionally, the proposed control can be applied directly online after initialization. Thus, the time-consuming offline training process is avoided. Furthermore, the proposed controller design also avoids the singularity problem, which may exist in regular feedback linearization based controls. Co-simulation using Matlab/Simulink and PSIM demonstrates the effectiveness of the proposed controller design.