Abstract: The traditional A^* algorithm applied to the path planning of offshore wind farm operation and maintenance ships has not yet taken into account the dynamic obstacles, water currents, and crossing navigation channels, therefore, this paper proposes a path planning method that considers navigational risks, heading angle constraints, and path smoothing. On the basis of constructing the map of offshore wind farm water environment by raster method, weight coefficients are introduced to change the proportion of estimated surrogate value in the total cost function of the A^* algorithm to achieve the purpose of balancing the strength of heuristic information and shortening the pathfinding time, and the risk of obstacles containing water currents is taken into account in order to improve the actual cost function of the A^* algorithm and enhance the security of the planned paths. Meanwhile, the heading angle constraint is considered in the A^* algorithm to reduce the total number of traversal nodes, the eight-neighborhood search is constrained to three neighboring nodes conforming to the path direction, the inflection points are extracted and visibility check is performed to remove the redundant inflection points in the path, and the smooth planning path is obtained using a uniform B-spline curve. Taking the Donghai Bridge No.5 and No.6 wind farm waters as an example, a high tide path planning scenario is established, and the operation and maintenance ship needs to pass through 9 wind turbines in order to complete the operation and maintenance tasks; 4 indexes（path length, total risk value of the path, total number of traversed nodes, and total number of inflection points） are utilized for evaluating the planning path, so as to validate the effectiveness of the improved A^* algorithm. The simulation results show that in the high tide scenario, the planning path smoothness of the improved A^* algorithm is improved by 77.69%, the total risk value of the planning path is reduced by 52.83%, and the total number of traversal nodes is reduced by 30.58%, but the planning path length of the improved A^* algorithm is 252.89 m longer than that of the traditional A^* algorithm.