Intelligent prediction method for permeability of layered phase controlled carbonate reservoirs based on BP neural network
Carbonate reservoirs are significantly impacted by diagenesis,exhibiting complex pore-throat structures and a low correlation between porosity and permeability.Predicting permeability in such reservoirs is challenging,and conventional methods relying on linear relationships often yield unsatisfactory results.This study proposes a comprehensive permeability prediction method based on BP neural network,which proves effective for layered phase controlled carbonate reservoirs.The method comprises three main steps.Firstly,quality control is applied to core and well log data.Subsequently,considering geological features,optimal parameters for predicting well log curves and the neural network model are selected to establish the prediction model.Finally,multiple data sources are integrated,and quality control is performed on the prediction results.The application of this method to carbonate reservoir A in the Middle East region yielded favorable permeability prediction results.Given the spatial complexity of carbonate reservoirs with well-developed pores,cavities,and fractures,the measured permeability of core plugs only represent local locations.In contrast,dynamic effective permeability measurements from well logs cover a larger range,reflecting the characteristics of the reservoir space.Additionally,with low clay minerals content and no sensitivity issues or significant anisotropy,the dynamic permeability values from well logs tend to be higher than those from core measurements.
万方数据
维普
