Urban inland inundation simulation grid adaptive optimization method based on KNN-GAT-DDQN

The invention discloses an urban inland inundation simulation grid adaptive optimization method based on KNN-GAT-DDQN, and the method comprises the steps: generating an initial high-precision grid in a research region through employing a grid generator, calculating the land utilization distribution and proportion in each grid according to the land utilization type, and obtaining the initial state S of the grid; performing feature classification by using a KNN algorithm; and calculating feature vectors h'i (K) of all N nodes by adopting a multi-head attention mechanism of a graph attention network, obtaining a grid feature vector # imgabs0 #, iteratively performing strategy selection on the grid feature vector # imgabs0 # by adopting an improved deep Q network, and outputting a final grid state and a node selection scheme Q (S, a). According to the method, the optimal strategy is selected based on the high-precision model grid, and high-efficiency adaptive grid optimization is realized. 本发明公开了一种基于KNN-GAT-DDQN的