Abstract: Taking Huangtukeng Coal Mine in Fujian Province as the research object, using network parallel electrical method technology, this paper expounds Levenberg Marquardt inversion principle and model construction from the perspective of three-dimensional detection data processing, and implements resistivity data inversion by introducing smooth constraint model and adaptive adjustment of damping factor, thus, the spatial distribution of apparent resistivity in different depth detection abnormal areas under the floor of the coal mining face are obtained. The results show that there are six relatively low resistivity anomaly areas in the detection area, and the water-richness is half strong and half weak, among which Dz1, Dz2 and Dz6 are weak; Dz3, Dz4 and Dz5 are strong, showing the characteristics of the water-richness of stronger in shallow than in deep, stronger in the north and weaker in the south, which is consistent with the drilling verification results and has high reliability.