Abstract: Due to their endurance enabled by wind energy utilization, unmanned sailboats are increasingly employed for marine monitoring and data collection. To tackle uncertainties in path planning and varying wind conditions during navigation, this study proposes a path planning algorithm based on dynamic wind field reconstruction and reinforcement learning. The method uses an onboard meteorological station to sample local wind data in real time and integrates a dynamic wind model reconstructed from the Navier–Stokes equations to estimate the global wind field, thereby enabling real-time path optimization for sailboats in dynamic environments. At the path planning level, an improved Q-learning framework is adopted, in which a state space incorporating both vessel position and real-time wind information is constructed. A reward function that comprehensively considers wind propulsion benefits and obstacle avoidance is designed, closing the loop between environmental perception and decision-making control. Simulations and experiments show that the proposed algorithm improves the efficiency of path planning, effectively reduces sailing time, and enhances the autonomous navigation and environmental adaptability of sailboats in complex marine environments.