Abstract: Identification of coal body structure by logging curves is an efficient and economical geophysical method. However, due to the influence of sedimentary environment and coal reservoir property, logging curves have multiple solutions, leading to an unclear logging response of coal body structure. And the identification rules acquired in one place cannot be applied in another place. Therefore, before starting logging discrimination, it is necessary to control the factors influencing well logging apart from coal structure. This essay takes an example from No.3 coal seam in Mabidong Block, Qinshui Basin. Firstly, coal cores are relocated to their logging depth by the positive correlation between ash content and gamma logging curves. And then, the coal-accumulating environment is divided by the ratio of vitrinite content to inertinite content. After that, the logging curves of similar environment are selected preferably. The result shows that resistivity logging serial curves can indicate the changes in coal body structure clearly, while acoustic wave logging cannot because of its shallow penetration depth. Finally, the Random Forest Model built by the chosen curves is used to predict other wells' coal body structure. The predicted results and the measured fracturing curves show that the results are in good agreement with the measured data. According to the application, the method can predicate the allocation of coal body structure, instruct hydraulic fracturing, which reduces development cost, and provide guidance to the logging response study on multi-regional coal body structure.