This study presents a thermo-mechanically coupled finite element framework with transient heat source modeling to predict welding residual stress distributions of Q355 steel butt joints, complemented by creep-incorporated analysis to evaluate post-weld heat treatment (PWHT) effects. The dual-ellipsoid heat source model demonstrated high accuracy (<2% thermal deviation, <10% geometric discrepancy) in capturing welding thermal dynamics and weld geometry. Residual stress distributions are longitudinally dominant, peaking at 379 MPa, with numerical predictions validated via blind-hole test (85% accuracy). Creep-integrated PWHT modeling yields a more accurate result relative to the modeling without considering creep, and the conclusion was validated by experiment. Elevated holding temperature or soaking duration promoted stress relaxation, while higher heating rates under fixed conditions inversely correlated with relaxation efficiency. The maximum welding residual stress could be relieved by nearly 80% with optimized PWHT process. These findings advance predictive welding mechanics and PWHT optimization strategies.