LOGGING EVALUATION OF SHALE GAS-BEARING PROPERTIES BASED ON LM-BP NEURAL NETWORK MODEL

Shale has complex macro and micro geological structures, and its internal clay mineral components mostly have the characteristics of bedding growth. Therefore, shale is also called "transversely isotropic medium". In this paper, based on an optimized neural network method, the gas-bearing properties of strongly heterogeneous shale are evaluated by logging. On this basis, we explore the influencing factors of gas-bearing properties of continental shale. The results show that when the BP neural network model is optimized by introducing the LM algorithm, the average relative error of using this method to predict the total gas content of shale reservoirs is about 5.52%. Factors affecting the gas-bearing properties of continental shale include organic matter content, mineral composition and content, petrophysical properties, gas occurrence state, and overpressure. Shale gas content has a good positive correlation with effective porosity and gas permeability. Shale with higher gas content has higher gas saturation. The adsorbed gas content of the study continental shale accounts for about 60%, which is mainly dominated by adsorbed gas. This indicates that the adsorbed state is an important gas storage mode in continental shale reservoirs.