Adaptive Model Predictive Control with Delay Compensation and Online DAB Gain Estimation for a Synergetic Isolated VIENNA Rectifier

The synergetic isolated VIENNA rectifier (SynC-VR-i) shapes its grid currents through two dual active bridges (DABs) fed from a split DC link of 10 µF per capacitor. Errors in the DAB power model therefore quickly appear as grid-current distortion. This paper proposes a delay-compensated adaptive model predictive control (MPC) scheme. A unit-invariant normalized-gradient estimator identifies the power gain of each DAB online, and the adapted model is used for both prediction and delay compensation. Under a positive-power excitation condition, the gain estimates are proven to converge exponentially to a neighborhood set by the power-model error. Local stability of the periodic switched operating point is estimated numerically along simulated orbits with the monodromy matrix, because frozen-time eigenvalues are unreliable here. The same evaluation yields a stability map over the horizon and the output-current weight that guides the tuning. In switching-level simulations of a 20 kW charger with 30% lower leakage inductance, the PI cascade and the non-adaptive MPC become unstable. The proposed controller keeps the grid-current total harmonic distortion at 2.3%. A compiled C implementation takes at most 5.1 µs per 10 µs control period on a desktop processor but 0.75–2.7 ms on a TMS320F28335 microcontroller, too slow for 100 kHz.