Predominant flow areas in fractured reservoirs, causing a large amount of fluid loss, are the primary reason for a failed practical enhanced geothermal system project. However, in practical situations, the existed geophysical exploration methods (e.g., microseismic techniques and tracer techniques) are unable to clear all the actual fracture network distribution in fractured reservoirs, so it is difficult to verify the existence state of predominant flow areas. This study aims to create an inversion method based on a fuzzy-genetic algorithm for identifying fracture network distributions, through discussing the influences of different initial guessed fracture network distributions on inversion results. Results show that the random distribution and the fracture aperture range have an obvious impact on inversion results (inversion accuracy rate = 70%). An increase in fracture length scope (>120−210 m) and number (>170) decreases inversion accuracy. Curved-fracture network distribution also leads to a decrease in the accuracy of inversion results. This study would provide suggestions for deep geothermal exploration.