Abstract: Abstract：Objectives To make local path planning algorithms more consistent with the maneuvering characteristics of ships, thereby generating safer and more reliable reference paths. Methods This paper proposes a three-dimensional potential field modeling method. By converting the Cartesian coordinate system to the ellipsoidal coordinate system, it addresses the anisotropy problem of the potential field distribution function. The ship’s potential energy distribution function is calculated by solving the Laplace equation. A control framework combining the potential field model and Model Predictive Control (MPC) algorithm is designed to enhance the adaptability of dynamic real-time local path planning for ships in different scenarios. Results Simulations were conducted with actual navigation vessels in the waters of the Sutong Yangtze River Highway Bridge area. The three-dimensional potential field model was used to obtain local reference paths, and MPC algorithm was employed for ship path tracking control simulation experiments. The results show that, compared to reference paths generated by traditional artificial potential field and improved artificial potential field methods, the three-dimensional potential field model’s local reference paths are superior in terms of length, curve smoothness, maximum steering angle, and average absolute heading error. This model can generate shorter and smoother local paths, which are more consistent with the actual maneuvering habits of ships, and exhibit less jitter in traffic-intensive scenarios. Conclusions The study demonstrates that the local reference paths generated by the three-dimensional potential field model can effectively identify the target ship’s steering angles and differences in ship scale, reduce reliance on the number of surrounding target ships, and effectively capture the interactions between ship agents, thus showing high reliability.