Abstract: To address the stability issues of DC microgrid buck converters under the coupled effects of model mismatch, complex noise, and constant power load （CPL） disturbances, a synergistic RAUKF-MPC control method is proposed in this paper. At the estimation layer, a robust adaptive unscented kalman filter （RAUKF） is constructed by integrating Huber robust weighting, Mahalanobis distance-based outlier detection, and adaptive noise covariance adjustment to suppress estimation biases caused by parameter perturbations and non-gaussian outliers. At the control layer, the recursive least squares （RLS） algorithm is employed for the online reconstruction of the predictive model, which is combined with the alternating direction method of multipliers （ADMM） utilizing an adaptive penalty factor to enhance the real-time solving capability of model predictive control （MPC）. Simulation results demonstrate that under pure gaussian noise and mixed noise conditions, the root mean square errors （RMSE） of voltage estimation are 0.0408 V and 0.0602 V, respectively. After model reconstruction, the open-loop state update error is significantly reduced from 0.58 V to 0.13 V. During the heavy load phase of the CPL, the average output voltage is maintained at 119.934 V, with an RMSE of 0.213 V and a voltage ripple of 1.55 V. The results indicate that the proposed method achieves high precision estimation and high-performance control even under the coexistence of multiple uncertain factors, providing an effective basis, technical support, and reference for the robust optimal operation of DC microgrid converters under complex operating conditions.