Abstract: To address the significant accuracy degradation in traditional evaporative waveguide inversion under non-ideal observational conditions—arising from ‘one-dimensional assumption distortion’ due to atmospheric horizontal inhomogeneity and extreme sensitivity to meteorological parameter perturbations—this study proposes an NPS-Transformer inversion algorithm integrating physical principles with deep learning. The method constructs a high-fidelity electromagnetic dataset incorporating gravity wave perturbations and multi-scale parameter drift characteristics based on the Wide-Angle Parabolic Equation (WAPE). The algorithm follows a “physically guided + data-corrected” strategy: first, designing a cross-modal attention module to deeply fuse meteorological priors with radar features; second, utilising Transformers to learn nonlinear residuals between physical benchmarks and true states; finally, introducing physical gradient constraints to suppress error amplification. Experiments demonstrate exceptional robustness in simulations featuring 30% systematic relative humidity bias and high atmospheric turbulence uncertainty: compared to traditional NPS models, the average absolute error (MAE) of corrected refractive index decreases by 86.7%, effective detection rate (i.e., detection probability) for low-altitude evaporative waveguide events rises to 96.8%, and electromagnetic blind zones are accurately reconstructed. This approach effectively overcomes the performance limitations of purely physical models when input data contains biases and disturbances. It achieves deep integration between meteorological and oceanographic information and waveguide physical mechanisms, providing a reliable solution for high-precision electromagnetic sensing in complex environments.