Prediction method of coal seam porosity based on BP neural network

Ping Zhou; Xinsi Hu; Danni Wei

doi:10.54097/jawsys62

Authors

Ping Zhou
Xinsi Hu
Danni Wei

DOI:

https://doi.org/10.54097/jawsys62

Keywords:

BP neural network, Coal seam porosity, Logging curve, Predictive model, Machine learning

Abstract

Coal seam porosity is a key parameter for evaluating the physical properties of coalbed methane (CBM) reservoirs, and its accurate prediction is crucial for CBM exploration and development. Traditional log interpretation methods exhibit limited accuracy when dealing with multifactor nonlinear relationships. This paper proposes a coal seam porosity prediction model based on an error back propagation (BP) neural network, which establishes the nonlinear relationship between optimized conventional logging curves (AC, CAL, CNL, DEN, GR, RT) and porosity (POR). Input variables were screened using Spearman correlation analysis, and a three-layer BP neural network structure was constructed. The network weights and thresholds were optimized via the back-propagation algorithm. Experimental results demonstrate that the model effectively learns the complex mapping relationship between logging parameters and porosity. The predicted results show strong agreement with measured values, with an average relative error of less than 8%, outperforming traditional regression methods. This approach provides a reliable technical means for predicting coal seam porosity.

Downloads

Download data is not yet available.

References

[1] Jin Zeliang, Xue Haifei, Gao Haibin, et al. Coalbed methane reservoir logging evaluation technology and application [J]. Coalfield Geology and Exploration, 2013, 41(02): 42-45.

[2] Yu Zhaolin, Wang Yanbin, Yu Yun, et al. Porosity logging prediction model of deep coal reservoir based on diversified regression [C]//The Petroleum Geology Committee of the China Petroleum Society, the Coal Methane Professional Committee of the China Coal Society, the Coal Methane Industry Technology Innovation Strategic Alliance. New Progress in Coal Methane Exploration and Development Technology - Proceedings of the 2018 National Coal Methane Academic Seminar. School of Geoscience and Surveying Engineering, China University of Mining and Technology (Beijing), 2018:424-429.

[3] Yang Liuqing, Chen Wei, Cha Bei. Research and application of prediction of reservoir porosity using convolutional neural networks[J]. Progress in Geophysics, 2019, 34(04):1548-1555.

[4] Shao Xianjie, Sun Yubo, Sun Jingmin, et al. Explanation method of coal rock parameter logging - Taking Hancheng Mining Area as an example [J]. Petroleum Exploration and Development, 2013, 40(05):559-565.

[5] Shen Yanjun, Ma Wen, Wang Xu, et al. Research on the correlation between pore development and intensity characteristics of oil-rich coal in different coal seams[J]. Coal Mine Safety, 2023, 54(03):161-168.

[6] Zheng Guiqiang, Yang Defang, Li Xiaoming, et al. Research on the characteristics of deep coal reservoir logging in the Shizhuang North Block in Qinshui Basin [J]. Coal Science and Technology, 2019, 47(06): 178-186.

[7] Zhang Chao, Huang Huazhou, Xu Delin, et al. Coal structure identification based on logging curves[J]. Coal Science and Technology, 2017, 45(09): 47-51.

[8] CAI C Z, LI G S, HUANG Z W, et al Experimental study of the effect of liquid nitrogen cooling on rock pore structure[J]. Journal of Natural Gas Science and Engineering, 2014, 21: 507-517.

[9] SALMACHI A, ZEINIJAHROMI A, ALGARNI M S, et al Experimental study of the impact of CO₂ injection on the pore structure of coal: A case study from the Bowen Basin, Australia[J]. International Journal of Coal Geology, 2023, 275: 104314.

[10] Hu Rui, Li Yong, Liu Yingtian, et al. Prediction method for logging reservoir parameters based on CL-Trans model [J]. Geophysical and chemical exploration computing technology, 2025, 47(03): 410-419.

[11] Liu Zhenming, Wang Yanbin, Han Wenlong, et al. Prediction of coal structure of BP neural network based on well logging data[J]. China Coal, 2018, 44(06): 38-41+55.

[12] Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323(6088), 533–536.

[13] Bergstra, J., & Bengio, Y. (2012). Random search for hyper-parameter optimization. Journal of Machine Learning Research, 13(Feb), 281–305.

[14] Box, G. E. P., Hunter, W. G., & Hunter, J. S. (1978). Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building. John Wiley & Sons.

[15] Hauke J, Kossowski T. 2011.Comparison of values of Pearson's andSpearman's correlation cefficients on the same sets of data. Quaestiones Geographicae, 30(2):87-93.

[16] Myers L, Sirois M J. 2014. Spearman correlation coefficients, differences. Between .// Wiey StatsRef:Statistics Reference Online.Washington: John Wiley & Sons.

[17] Nair, V., & Hinton, G. E. (2010). Rectified linear units improve restricted boltzmann machines. In *Proceedings of the 27th International Conference on Machine Learning (ICML-10)* (pp. 807–814).

[18] Qian, N. (1999). On the momentum term in gradient descent learning algorithms. Neural Networks, 12(1), 145–151.

[19] James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An Introduction to Statistical Learning: with Applications in R. Springer.