<p>Modern cross-regional remote state estimation is essential for secure and stable smart grid operation. However, cyberattacks introduce substantial security risks, and model uncertainty arising from inaccurate parameterization as well as linearization approximation errors in nonlinear dynamics makes it difficult for conventional estimation methods to simultaneously satisfy stringent requirements on accuracy and robustness. To address these challenges, this paper proposes a remote robust state estimator designed for the coexistence of DoS attacks and model uncertainty. The proposed method explicitly accounts for DoS-induced data losses and incorporates a sensitivity penalization-based robust correction into the estimation cost function to mitigate performance degradation caused by unmodeled dynamics and linearization errors, thereby enabling recursive remote estimation with improved accuracy and enhanced robustness. We further prove that the resulting pseudo-one-step prediction error covariance matrix associated with the proposed estimator satisfies a Riccati recursion, and we validate the effectiveness of the proposed approach on the Kundur two-area, four-generator system. Simulation results demonstrate that, compared with representative baseline methods, the proposed estimator substantially reduces estimation errors and improves estimation accuracy and robustness under the combined effects of DoS attacks and model uncertainty, while maintaining reliable real-time recursive performance.</p>