A three-dimensional heterogeneous
aquifer horizontal equivalent
permeability coefficient multi-scale prediction method based on a residual network comprises the steps that a foundation pit
dewatering two-dimensional pressure-bearing steady flow model is constructed, the influence of the size of a lenticular body on foundation pit
dewatering flow field disturbance is evaluated, and the resolution ratio of a parameter prediction model grid is determined; based on actual drilling data and transition probability /
Markov chain T-Progs, according to multiple deposition types formed by a binary deposition structure, a multi-layer
aquifer structure, a coarse-grained deposition structure and a fine-grained deposition structure, a three-dimensional stratigraphic
random structure model set is constructed, and three-dimensional
stratigraphic section random slices are established; obtaining horizontal equivalent permeability coefficients corresponding to stratigraphic sections in various stratigraphic structures based on the underground
water flow numerical model FloPy, and establishing a stratigraphic structure-equivalent
permeability coefficient training set based on a neural network; a
deep learning model is constructed based on the residual network ResNet50; and training, verifying and testing the equivalent
permeability coefficient of the stratum section formed by the binary sedimentary structure, the multi-layer
aquifer structure, the coarse grain sedimentary structure and the
fine grain sedimentary structure.