Abstract: The full waveform inversion (FWI) is a core technology for high-precision velocity modeling. However, conventional least-squares diving-wave FWI is limited by insufficient inversion accuracy and depth, failing to meet the velocity modeling needs from deep-water complex fault zones. Taking phase-driven FWI as the key issue,high-precision velocity modeling under deep-water complex fault conditions was conducted. At algorithm level, the FWI objective function was optimized to enhance the utilization of P-wave phase information and reduce FWI’s dependence on the amplitude of low-frequency diving waves, improving inversion stability and effective depth. At data level, key acquisition parameters (e.g., streamer length, trace interval) were optimized to improve fold coverage and acquisition density, by which the adaptation between acquisition geometry and FWI was realized. The results were then verified via numerical simulations and seismic data, showing that phase-driven FWI could effectively reduce the dependence of traditional diving-wave FWI on spread length and improve the velocity inversion accuracy. Moreover, optimization tests on the acquisition geometry indicated that, appropriately reducing the trace interval could enhance the FWI inversion accuracy, while reasonably increasing the spread length could improve the stability of deep velocity inversion. This study demonstrated the technical advantages of phase-driven FWI and the design principles of matched acquisition geometry for deep-water complex fault zones, and provided a reliable technical support and theoretical reference for high-precision velocity modeling in similar areas.