Abstract: Accurate and rapid prediction of meniscus flow velocity under different process conditions is crucial for preventing slag entrapment in continuous casting. Due to the complexity of high-temperature, multi-physics coupling, real-time measurement of the mold flow field in mold is challenging. The paper develops a numerical model coupling mold melt flow and SEN argon blowing. Based on precise simulations, a fitting formula is proposed to predict the maximum meniscus velocity using liquid level differences. To achieve real-time meniscus velocity prediction, a GA-SVM model is developed, which integrates genetic algorithms, support vector machines, and numerical simulations using 24 sets of simulation data. The GA-SVM model achieves an R² of 0.93, providing high accuracy while avoiding complex numerical calculations. It enables rapid velocity prediction while avoiding complex simulations and serves as a surrogate model for digital twin applications in continuous casting.