Temperature-induced drift in the winding resistance of long-stator linear synchronous motors (LSLSMs) critically affects the levitation control performance of high-speed magnetic levitation (maglev) trains. To address this issue, this paper proposes a collaborative optimization framework that integrates multiphysics coupling analysis with active disturbance rejection control (ADRC). First, an electromagnetic-thermal-fluid coupling model of the LSLSM is developed through finite-element co-simulation. Second, a fast temperature prediction model is established by combining the parameter fitting with a lumped parameter thermal network (LPTN), enabling dynamic estimation of the winding temperature and the associated resistance drift. Third, an extended state observer (ESO) is employed to estimate and compensate for thermally induced parameter drift in real time. Numerical simulations under multiple operating scenarios show that the proposed method achieves improved tracking performance, stronger disturbance rejection capability, and better robustness against thermal-induced parameter drift than the benchmark proportional-integral-derivative (PID) controller. In addition, considering thermal-induced parameter drift in the controller design process leads to more consistent closed-loop performance in the levitation plant.