Abstract
In this paper an adaptive sliding-mode model predictive control (ASMPC) scheme is proposed for precise regulation of a low-voltage, non-inverting buck-boost converter's output voltage. By embedding a sliding-mode term within a constrained MPC framework and enforcing explicit stability criteria, the controller accelerates convergence and enhances closed-loop robustness. A linear state-space model is updated online via a gradient-based Levenberg-Marquardt algorithm to compensate for converter nonlinearities and reduce reliance on an offline model. Simulation results in a photovoltaic system and experimental tests on a DC/DC converter confirm that ASMPC maintains tight voltage tracking under varied reference voltages, load changes, and input-voltage fluctuations. Compared with baseline adaptive MPC and conventional PI control, ASMPC achieves faster settling, smaller voltage ripples, and energy-efficiency improvements of approximately 3%-6% across all scenarios.