Abstract: To address issues such as low weld seam deviation recognition accuracy and poor adaptability of oscillation control in the multi-layer, multi-pass welding process of medium-thick plates for wind power towers, a comprehensive approach is adopted, integrating theoretical analysis, simulation, experimental validation, environmental construction, and actual production application. This includes process feature extraction and modeling of four-wire narrow-gap submerged arc welding (SAW), hybrid seam tracking using arc and laser vision sensing, multi-layer tracking in narrow-gap SAW, and closed-loop real-time control.The research is specifically focused on developing a fully automatic four-wire narrow-gap SAW seam tracking technology that requires no teaching, a mathematical model that is capable of real-time compensation for changes in welding parameters, and dual-sensor integration of arc and visual sensing. Experiments are conducted on 80 mm-thick Q420MC steel plates using four-wire SAW. The weld surfaces are found to be uniformly formed, and macro-metallographic analysis shows no cracks, inclusions, or other defects. Welding efficiency is improved by 300%, and 100% of non-destructive tests are passed, providing effective technical support for the intelligent manufacturing of wind power tower structures.