Building fires are characterised by rapid fire spread, high smoke concentration, low visibility, etc., which makes evacuation of people extremely difficult. Traditional path planning algorithms have many limitations when facing complex dynamic environments such as fires, and cannot meet the demand for real-time, efficient and accurate evacuation. In order to overcome these problems, this paper proposes a fire evacuation path dynamic planning system based on the improved A* algorithm. The system combines the dynamic characteristics of the fire scene and the efficient path search capability of the A* algorithm to plan the optimal evacuation path in real time. Firstly, Pyrosim software is used to simulate the fire spreading process in the building and analyse the fire danger of each time period according to the simulation results; then the escape paths of the evacuees in the building are dynamically planned according to the fire spreading law of the building, and through the comparative experiments, the results show that the number of expansion nodes of the improved A* algorithm is reduced by about 87%, and the speed of obtaining the distance of the paths is increased by about 80%; finally The improved A* algorithm releases a dynamic real-time evacuation visualisation scene relying on Unity 3D to provide escape guidance for evacuees. The system significantly improves fire rescue efficiency and personnel safety by simulating fire scenes, analysing fire changes in real time, and dynamically planning optimal evacuation paths.