Laser Powder Bed Fusion (LPBF) is a promising technology for fabricating intricate three-dimensional structures. Efforts have been made to enhance the properties of fabricated parts by controlling material microstructures through tailored scanning strategies [IOP Conf. Ser.: Mater. Sci. Eng., 1310(1) (2024) 012013]. However, given the vast number of possible scanning strategy combinations, numerical simulations for microstructure prediction are essential. For predicting grain-scale microstructures in LPBF, the cellular automaton (CA) method has been widely employed. In our previous study, we demonstrated that the multi-phase-field (MPF) method enables microstructure prediction that accounts for scanning strategies involving multiple layers and multiple laser tracks, leveraging high-performance parallel computing on multiple GPUs. The MPF approach provides smooth interfaces and detailed three-dimensional microstructures, offering advantages over conventional CA methods. In this study, we used Rosenthal’s equation—which provides an analytical solution for the temperature distribution induced by a moving point heat source—to model the temperature field generated by laser scanning and incorporated it into the MPF method [Mater. Trans., 64(6) (2023) 1150-1159]. However, this approach has limitations, particularly at relatively high laser powers, where the melt pool geometry deviates from the idealized semicircular shape expressed in Rosenthal’s solution. To address this, we developed a method that bridges thermal-fluid simulation and MPF simulation to enable microstructure prediction under various melt pool conditions in LPBF, including those at high laser power. Using the proposed method, we conducted microstructure prediction simulations for a range of scanning strategies. The resulting microstructures were evaluated from multiple perspectives, including two-dimensional cross-sections, three-dimensional morphologies, and grain size distributions.