With the aim of minimizing task completion time and reducing resource consumption, this paper investigates the dynamic task allocation problem of a heterogeneous aircraft cooperative cluster with multiple fire spots in the forest. Firstly, we establish the fire point propagation model and task assignment model to enhance task assignment accuracy. Based on this, we propose the Chaotic Cosine Fusing T-Distribution Variation and Quasi-Reflection Learning Grey Wolf Optimization algorithm (CTQGWO), which integrates distributed mutation and quasi-reflective learning. By incorporating chaos population initialization of the Tent map, adaptive control parameter adjustment strategy, distributed mutation, and quasi-reflective learning strategy, we can enhance the convergence speed and accuracy of the algorithm. Finally, taking into account the complexity of the forest and the instability of the aircraft, we develop a task reassignment algorithm that can quickly allocate and reduce the number of communications. Through experiments, our algorithm enables rapid decision-making for forest fire emergency scheduling, under the condition of limited forest fire resources, effectively reducing the total rescue time and minimizing the consumption of rescue resources.