Abstract：
The pore types of Carboniferous KT-II carbonate reservoirs in the middle block of the eastern margin of the Caspian Basin are complex and variable, which makes it difficult to predict the permeability accurately. In the studied area, application analysis was undertaken of six permeability prediction models based on capillary pressure curves. These were the Purcell model, Swanson parameter, capillary-parachor parameter, R25, R35 and R50. On this basis, seven sensitive parameters including porosity and other parameters obtained from capillary pressure curve are selected as input vectors, and the particle swarm optimization support vector machine (PSO-SVM) method is used to predict permeability. Results show that the traditional permeability models provide unreasonable results. Although the Purcell model is superior to other models, the coefficient of determination between predicted and measured results is only 0.763. The permeability predicted by support vector machine is reliable. The coefficients of determination of predicted results of training samples and test samples with measured values are 0.913 and 0.854, respectively. The proposed method overcomes the shortcomings of the traditional permeability model to carbonate reservoirs to a certain extent, and provides valuable information for formation evaluation.
Key words：carbonate reservoir; permeability prediction; capillary pressure curve; support vector machine
Corresponding Authors： maozq@cup.edu.cn
Cite this article：ZHAO Peiqiang, LI Changwen, SHA Feng, ZHANG Lili, MAO Zhiqiang, JIANG Xinyu. Study of permeability prediction of carbonate reservoirs in the middle block of the eastern margin of the Caspian Basin. Petroleum Science Bulletin, 2020, 01: 39-48.
