With the development of modern agriculture, the level of agricultural automation and machinery automation has been significantly improving, and an increasing number of intelligent algorithms have been applied to the cultivation, management, and harvesting processes in agriculture. In this research, we introduces an innovative task planning algorithm for unmanned aerial vehicle (UAV) swarms conducting plant protection tasks, which can enhance the precision and efficiency of spraying operations in smart farm scenarios. The algorithm utilizes a pointer network (Ptr-Net) integrated with an actor-critic (AC) reinforcement learning approach, enabling adaptive responses to real-time changes in job demands and operational environments. Simulation experiments conducted under 20 UAVs and 10 target areas demonstrate the algorithm’s superior performance in execution speed and task completion rate compared to random strategy and traditional methods. The findings highlight the efficacy of neural network applications in agricultural settings, contributing to the progress of precision farming technologies.