Identificating favourable reservoir areas of coalbed methane based on multifractal and gated circulation unit

Abstract

Identificating favourable reservoir areas of Coalbed methane (CBM) is significant for improving economics of CBM project. However, accurately identification of favourable areas from CBM reservoir with strong heterogeneity of reservoir parameters is challenging. In previous researches, some traditional models were applied to identify favourable areas of CBM reservoir, however there is a widely gap between the evaluation result and reality distribution of high-yield areas of CBM reservoir. For solving above problem, identifying favourable reservoir area with strong heterogeneity is conducted with deep learning and multifractal theory. In identification process, firstly the fracture characteristics of the research area is calculated with multifractal theory to generate weighted data layer. Secondly a deep learning model is constructed with the genetic algorithm and door circulation unit to identify favourable reservoir areas of CBM and is verifyed with the block of Fanzhuang-Zhengzhuang located in the Qinshui coalfield of China. The fitting degree of the above model and the known CBM wells is better than the previous methods, and the accuracy of the model reaches 87%, which indicates that the method is feasible and provides a new way for identificating favourable reservoir areas of CBM under complex geological conditions.