Three-dimensional path planning is one of the key technologies for UAVs to achieve autonomous movement. Due to the influence of environmental diversity, currently commonly used algorithms have certain limitations. This paper proposes a three-dimensional planning algorithm for UAVs based on thermal gradients. This algorithm uses the idea of solving the heat transfer path in the steady-state thermal potential field to analogize the UAV’s flight path, and regards the UAV’s optimal path as a stable The path with the fastest temperature drop in the state thermal potential field, the key navigation points planned according to the principle of the fastest drop in thermal gradient, and the trajectory after path planning were optimized, and a safe and feasible trajectory was successfully planned in the three-dimensional environment. The transformed thermal potential field contains global environmental information, and combined with the principle of the fastest temperature gradient descent search, it can overcome the blindness in the path planning process. There are no local minima in the thermal potential field, which allows the method to efficiently handle scenes with complex obstacles. The simulation and comparison results show that the algorithm has faster convergence speed compared to commonly used UAV path planning algorithms and is widely applicable to various three-dimensional environments.