The application of model predictive control (MPC) in single-inductor multiple-output (SIMO) DC-DC converters can effectively address the cross-regulation problem. However, the high sensitivity of MPC to model inaccuracies and parameter mismatch limits its practical adoption. To overcome this limitation, this article introduces a deep neural network based model predictive control (DNN-MPC) method that alleviates the dependency of MPC on precise system models. The proposed approach has been validated using a MATLAB/Simulink simulation model. The results indicate that DNN-MPC not only reduces energy consumption by   reducing switching actions but also delivers superior dynamic performance, tracking accuracy, and robustness against varying system parameters and operating conditions. Compared to the conventional MPC method, DNN-MPC achieves a 31% reduction in   settling time, smaller voltage ripples, and a 57% decrease in the number of switching actions, resulting in   lower energy losses. These simulation results affirm the effectiveness of DNN-MPC for the SIMO DC-DC converter.

deep neural network, model predictive control, single-inductor multiple-output dc-dc converter