With the increasing scale of forest fires and the increasingly prominent problems of slow response and low resource scheduling efficiency of traditional fire-fighting means, existing intelligent perception and control technology methods have problems such as decision lag, low coordination efficiency, and extensive energy consumption management when dealing with large-scale and dynamically evolving fire environments. To this end, this paper introduces the integration of artificial intelligence and Internet of Things technologies to construct a UAV collaborative operation scheduling model based on multi-agent deep reinforcement learning. In terms of artificial intelligence, the DDPG (Deep Deterministic Policy Gradient) algorithm is used to achieve optimal control of the UAV path, and the priority experience playback mechanism is combined to improve the strategy convergence speed and adaptability in dynamic fire scenes; in terms of the Internet of Things, by deploying a multi-node fire sensor network and edge computing terminals, high-frequency collection and real-time feedback of fire information are achieved, providing UAVs with global perception support and dynamic rescheduling capabilities. The experimental results show that the fusion model proposed in this paper is superior to the traditional scheduling method in key indicators such as fire response time, fire extinguishing resource allocation efficiency, and fire extinguishing energy consumption per unit area. When the number of fire sources is 3, the perception accuracy of the fusion method is 96.2%, which is 7.7% points higher than the 88.5% of traditional scheduling, significantly improving the system efficiency of large-scale forest fire extinguishing.