# Improved pore structure characterization and classification of strong diagenesis sandstones by data-mining analytics in Tazhong area, Tarim Basin

## TL;DR

This paper introduces a data-mining approach to classify pore structures in strongly diagenetic sandstones, improving understanding of reservoir properties in the Tazhong area.

## Contribution

A novel data-mining pipeline using clustering and PCA to classify pore structures in strongly diagenetic sandstones is proposed.

## Key findings

- A classification scheme using MICP data and data-mining analytics was established for pore structures in strong diagenesis sandstones.
- Three variable groups were identified through hierarchical clustering and PCA for pore structure characterization.
- Six classification schemes were validated using decision tree and LDA models, aiding in predicting oil and gas distribution.

## Abstract

The Silurian system in Tazhong area is characterized by extensive, low-abundance lithological reservoirs with strong diagenesis, resulting in significant heterogeneity. The complex pore structure in this area significantly impacts fluid control, making accurate characterization and classification of pore structures crucial for understanding reservoir properties and their influence on oil and gas distribution. Based on 314 Mercury Injection Capillary Pressure (MICP) samples in combination with core slices and thin casting slices observation, a pipeline of characterization and classification scheme by data-mining analytics of strong diagenesis sandstone pore structure types in the study zone is established, and the characteristics of different pore structures are clarified. According to the pore structure parameter abstracted by MICP data compression and variable analysis based on hierarchical clustering and principal component analysis (PCA) analysis, the variables are reasonably evaluated and screened, and the screened variables can be divided into three groups: mean pore throat radius-maximum pore throat radius-median pore throat radius-pore throat diameter mean variable group, microscopic mean coefficient variable group, and median pressure displacement pressure-relative sorting coefficient variable group. The combination of classification schemes analysed by decision tree model and linear discriminant analysis (LDA) model was determined. In the two-dimensional projection diagram of LDA model, a relatively obvious distribution of low displacement pressure, middle displacement pressure and high displacement pressure was obtained, and three distribution lines were nearly parallel. Based on the relevant information, 6 combined classification schemes suitable for final pore structure modelling were determined verified by microscopic observation. The correct characterization and classification of pore structure can be applied to the prediction of pore type, which can be used to improve the prediction of oil and gas distribution and oil and gas recovery in the future.