Research On Reservoir Pore Structure Evaluation And Reservoir Classification Prediction Method Based On Deep Learning

Accurately obtaining information on reservoir pore structure is very important for the exploration and development of complex oil and gas reservoirs,and is one of the key basis for effective reservoir classification.There are two main types of pore structure characterization methods: one is to use logging data to evaluate the pore structure;the other is to build an evaluation model based on physical experiments.There are large differences in the scale and resolution of these two types of methods.Data mining and deep learning algorithms have great application potential in small sample data modeling and prediction,and can provide new research ideas and methods for multi-scale data fusion.In this paper,data mining algorithms such as gray correlation analysis,principal component analysis,factor analysis,and hierarchical clustering are used to analyze mercury intrusion pore structure data to determine the main types of reservoirs;then,the reservoir types obtained by cluster analysis are used as label samples The data is applied to conventional logging data models,and fully connected neural networks and convolutional neural network models are used to predict reservoir types.The research results show that the reservoir type in the study area should be divided into 5 categories.Data mining and deep learning algorithms can be used to predict the reservoir type,which can achieve 82.1% prediction accuracy on the test set.When the training model is applied to test wells,the prediction results obtained are consistent with the reservoir porosity and permeability characteristics reflected by logging interpretation.This provides a new idea for the use of core analysis data and logging data for reservoir classification evaluation,and is of guiding significance for the prediction of oil and gas distribution and improving development production.