A hidden water-bearing collapse column (WBCC), directly exposed during coal mining, is a potential source of serious mine flooding. Presently, the transient electromagnetic method (TEM) is widely used, which shows good performance in detecting WBCCs in coal mines. However, traditional numerical simulation approaches have failed to be adapted to complicated underground environments affected by roadways, and the application of boundary conditions require a large grid. Considering this, the authors chose the convolutional perfectly matched layer (CPML) as a boundary condition, whose effects were found to be superior to the traditional Dirichlet boundary condition. Then, based on stratigraphic data related to coal measures located in North and Mid-eastern China, the whole-space geoelectric model was established. The whole-space TEM response of the WBCC at different depths of the stope face floor and ahead of the driving face were also simulated by the finite-difference time-domain method. The numerical simulation results indicate that the underground roadway exerted a substantial influence in early periods, yet the effects became negligible with time, according to the inductive potential decline curve. In addition, the contour distribution of apparent resistivity for different geoelectrical models of the WBCC was consistent with the models. The closer the roof of the collapse column was to the coal seam floor, the lower the apparent resistivity. Moreover, TEM was applied to hidden water-bearing collapse column detection in real underground coal mines.