Aiming at the difficulties of real-time decision-making and safe flight in low-altitude dynamic obstacle avoidance for electric vertical take-off and landing vehicles (eVTOL), a low-altitude path obstacle avoidance algorithm based on dynamic window-3-dimensional velocity obstacle (3DDWA-VO) fusion is proposed. The algorithm establishes a synergistic optimisation mechanism between kinematic constraints and dynamic avoidance by fusing the linear velocity constraint space of the dynamic window method and the predictive domain of the obstacle avoidance cone of the 3DDWA-VO method, so as to realise the double enhancement of the decision-making efficiency of the obstacle avoidance and the flight stability within the boundary of the safe velocity. Simulation results show that the paths planned by the proposed method can effectively complete the obstacle avoidance planning in single-dynamic obstacle and multi-dynamic obstacle environments. In a typical urban scenario with cross-motorised obstacles, the 3DDWA-VO algorithm significantly improves the success rate of obstacle avoidance compared to the traditional 3D velocity obstacle method, while the length of the obstacle avoidance path is reduced by 3.3%, and the planning time is shortened by 0.03 s. The method also optimises the mean curvature of the dynamic obstacle avoidance trajectory to 0.0136 rad/m, which effectively reduces the frequency of the vehicle's attitude adjustment, and provides real-time response capability and high reliability. response capability and high reliability, providing an effective solution for safe autonomous obstacle avoidance in complex dynamic environment of eVTOL.