Machine learning-based analysis of factors controlling the reservoir productivity of the Maokou Formation, Sichuan Basin

Abstract

By 2024, the proven natural gas reserves of the Maokou Formation in the central Sichuan Basin had exceeded 1.5×10¹¹ m³, suggesting great potential for hydrocarbon exploration. However, this formation exhibits small reservoir thicknesses and pronounced lateral heterogeneity due to sedimentation and erosion. These characteristics lead to significantly varying single-well production, complicating the identification of major controlling factors in high production and reducing the prediction accuracy of sweet spots. To identify the key geological factors controlling the gas production in the Maokou Formation, this study investigated the geological and seismic characteristics of the dolomite reservoirs in this formation. The primary geological factors influencing single-well production were explored using a machine learning approach—an enhanced random forest (RF) prediction model based on Shapley additive explanations (SHAP) values (also referred to as the SHAP-enhanced RF model). Accordingly, an optimal combination of geological parameters for high production was determined, followed by the identification of sweet spots. The results demonstrate that the SHAP-enhanced RF model allows for the effective quantification of the relative importance of various factors. The most critical factors affecting the gas production in the Maokou Formation include burial depth, natural gamma-ray value, the spatial distance of bright spots (defined as the time thickness of bright spots relative to the base of the second member of the Maokou Formation), and paleogeomorphology. High-yielding wells typically feature a joint advantage of multiple critical factors. A strong synergistic effect is prone to occur when several optimal factors fall within their optimal ranges simultaneously. In this case, high production capacity might be achieved. Compared to an equal-weight RF model, the SHAP-enhanced RF model yielded a mean absolute error (MAE) decreasing by 18.6%, thereby enhancing prediction accuracy and reliability. The findings of this study can serve as a valuable guide for production planning and well placement optimization in the Maokou Formation.

Keywords

Natural gas / Exploration potential / Dolomite reservoir / Sweet spot / High-yielding well / Earthquake parameter / Sensitivity-based factor selection / Shapley additive explanations (SHAP) values Production prediction of the Maokou Formation Sichuan Basin

References

[1] Chen YY, Zhao LX, Pan JG, Li C, Xu MH, Li KJ, Zhang FS, Geng JH. 2021. Deep carbonate reservoir characterisation using multi-seismic attributes via machine learning with physical constraints. Journal of Geophysics and Engineering, 18(5), 761-775. doi: 10.1093/jge/gxab049.

[2] Eftekhari M, Khashei-Siuki A. 2025. Evaluating machine learning methods for predicting groundwater fluctuations using GRACE satellite in arid and semi-arid regions. Journal of Groundwater Science and Engineering, 13(1), 5-21. doi: 10.26599/JGSE.2025.9280035.

[3] Guo TL, Zhu LT, Liu YT. 2024. Key technologies for accurate potential tapping in the mid-to-late stage of stable production of Changxing Formation bioherm high-sulfur gas reservoirs with bottom water in the Yuanba Gas Field, Sichuan Basin. Natural Gas Industry, 44(11), 72-81 doi: 10.3787/j.issn.1000-0976.2024.11.007. (in Chinese with English abstract).

[4] He X, Tang QS, Wu GH, Li F, Tian WZ, Luo WJ, Ma BS, Su C. 2023. Control of strike-slip faults on Sinian carbonate reservoirs in Anyue gas field, Sichuan Basin, SW China. Petroleum Exploration and Development, 50(6), 1116-1127 doi: 10.11698/PED.20220611. (in Chinese with English abstract).

[5] Huang DY, Lian ZH, Yan G, Ye K, Tao L, Zhang J, Halifu M, Cao ZX, Deng XA, Zhao YH. 2023. A new approach for multi-fractured horizontal wells productivity prediction in shale oil reservoirs. International Petroleum Technology Conference, Bangkok, Thailand, March 2023. doi: 10.2523/IPTC-23019-EA.

[6] Huang SP, Jiang QC, Jiang H, Tang QS, Zeng FY, Lu WH, Hao CG, Yuan M, Wu Y. 2023. Genetic and source differences of gases in the Middle Permian Qixia and Maokou formations in the Sichuan Basin, SW China. Organic Geochemistry, 178, 104574. doi: 10.1016/j.orggeochem.2023.104574.

[7] Hui G, Chen ZX, Wang YJ, Zhang DM, Gu F. 2023. An integrated machine learning-based approach to identifying controlling factors of unconventional shale productivity. Energy, 266, 126512. doi: 10.1016/j.energy.2022.126512.

[8] Jiang XM, Yan WC, Xing HL, Feng P, Sun JM, Fan YC. 2025. A highly accurate and interpretable gradient boosting machine learning model for predicting permeability in strongly heterogeneous reservoirs using the Optuna framework. Journal of Geophysics and Engineering, 22(3), 810-823. doi: 10.1093/jge/gxaf039.

[9] Kuang MZ, Zhang XB, Yuan HF, Chen C, Zhang XH, Peng HL, Xu T, Xiao QR, Li TJ, Shan SJ. 2024. Sequence stratigraphy and sedimentary facies characteristics of the Maokou Formation carbonate rocks in central Sichuan Basin. Journal of Palaeogeography, 26(5), 1201-1220 doi: 10.7605/gdlxb.2024.05.078. (in Chinese with English abstract).

[10] Liao Y, Zhang JY, Lu PD, Li ZQ, Li WZ, Tian TZ, Wu J, Sun W, Liu SG, Deng B. 2023. Multi-stage fluid activity and accumulation process of the Middle Permian Maokou Formation in the northern slope of central Sichuan Basin. Natural Gas Geoscience, 34(11), 1927-1940 doi: 10.11764/j.issn.1672-1926.2023.06.011. (in Chinese with English abstract).

[11] Liu JW, Zhao H, Wang Y, Zhang WJ, Chen K, Di ZX. 2022. Distribution characteristics and hydrocarbon geological significance of Permian volcanic rocks in Longhuichang-Longmen areas, eastern Sichuan Basin, SW China. Natural Gas Geoscience, 33(3), 381-395 doi: 10.11764/j.issn.1672-1926.2021.08.010. (in Chinese with English abstract).

[12] Qi Y, Li JX, Chen Q, Tu ZY, Zhao GX, Han XL. 2024. Evaluating the dominant factors affecting shale oil productivity based on machine learning: A case study of West Area 233, Qingcheng Oilfield. Drilling and Production Technology, 47(6), 61-68 doi: 10.3969/J.ISSN.1006-768X.2024.06.08. (in Chinese with English abstract).

[13] Shapley LS. 1951. Notes on the N-Person Game — II: The Value of an N-Person Game. RAND Corporation. doi: 10.7249/RM0670.

[14] Tan XC, He RY, Yang WJ, Luo B, Shi JB, Zhang LJ, Li MD, Tang YX, Xiao D, Qiao ZF. 2025. Origin and distribution model of thin dolomite reservoirs in the lower sub-member of Mao 2 Member of Middle Permian Maokou Formation in Wusheng-Tongnan area, Sichuan Basin, SW China. Petroleum Exploration and Development, 52(1), 112-127 doi: 10.1016/S1876-3804(25)60009-7. (in Chinese with English abstract).

[15] Tang HQ, Zhang BJ, Xiong XJ, Li TJ. 2024. Seismic prediction technology for complex beach facies dolomite reservoirs of Maokou Formation in central and northern Sichuan Basin: A case study of Jiaotan1 well area. Acta Petrolei Sinica, 45(9), 1385-1398 doi: 10.7623/syxb202409006. (in Chinese with English abstract).

[16] Tang QS, Liu JW, Wu GH, Tang S, Tian WZ, Li CH, Li SY, Huang TJ. 2024. Seismic description and application of deep carbonate strike-slip fault-controlled “sweet spots” reservoirs in the Anyue Gas Field, Sichuan Basin. Natural Gas Geoscience, 35(11), 2053-2063 doi: 10.11764/j.issn.1672-1926.2024.05.012. (in Chinese with English abstract).

[17] Tong SK, Wang FY, Gao HH, Zhu WY. 2024. A machine learning-based method for analyzing factors influencing production capacity and production forecasting in fractured tight oil reservoirs. International Journal of Hydrogen Energy, 70, 136-145. doi: 10.1016/j.ijhydene.2024.05.036.

[18] Wang Z, Tang HM, Cai H, Hou YW, Shi HF, Li JL, Yang T, Feng YT. 2022. Production prediction and main controlling factors in a highly heterogeneous sandstone reservoir: Analysis on the basis of machine learning. Energy Science and Engineering, 10(12), 4333-4923. doi: 10.1002/ese3.1297.

[19] Xie JP, Qin Y, Cao H, Zhang XH, Peng HL, Chen C, Gao ZL. 2025. Reservoir prediction method and effect analysis based on reflection coefficient decomposition: A case study of the Maokou Formation in central region of the Sichuan Basin. Natural Gas Geoscience, 36(3), 519-532 doi: 10.11764/j.issn.1672-1926.2024.09.012. (in Chinese with English abstract).

[20] Xie WR, Wen L, Wang ZC, Hao Y, Xin YG, Wu SJ, Li WZ, Yao QY, Ma SY, Chen YN. 2024. Types of unconventional marine resources and favorable exploration directions in the Permian-Middle Triassic of the Sichuan Basin. Natural Gas Geoscience, 35(6), 961-971 doi: 10.11764/j.issn.1672-1926.2023.11.013. (in Chinese with English abstract).

[21] Xu SY, Zhu Y, Zeng Y, Xiao XW, Lin Y, Zhao CN, He KL, Li Y, Li YW. 2024. Pore structure characteristics and classification evaluation of Maokou Formation reservoir in Penglai Gas Field, Sichuan Basin. Natural Gas Geoscience, 35(12), 2132-2141 doi: 10.11764/j.issn.1672-1926.2024.04.028. (in Chinese with English abstract).

[22] Xue T, Huang TJ, Cheng LB, Ma SW, Shi JC. 2021. Dominating factors on shale oil horizontal well productivity and development strategies optimization in Qingcheng Oilfield, Ordos Basin. Natural Gas Geoscience, 32(12), 1880-1888 doi: 10.11764/j.issn.1672-1926.2021.11.002. (in Chinese with English abstract).

[23] Yang Y, Xie JR, Zhao LZ, Huang PH, Zhang XH, Chen C, Zhang BJ, Wen L, Wang H, Gao ZL, Shan SJ. 2021. Breakthrough of natural gas exploration in the beach facies porous dolomite reservoir of Middle Permian Maokou Formation in the Sichuan Basin and its enlightenment: A case study of the tridimensional exploration of Well JT1 in the central-northern Sichuan Basin. Natural Gas Industry, 41(2), 1-9 doi: 10.3787/j.issn.1000-0976.2021.02.001. (in Chinese with English abstract).

[24] Yang YM, Wen L, Zhang XH, Chen C, Chen K, Zhang Y, Di GD, Wang H, Xie C. 2021. New exploration progress and prospect of Middle Permian natural gas in the Sichuan Basin. Natural Gas Industry, 8(1), 35-47 doi: 10.1016/j.ngib.2020.07.002. (in Chinese with English abstract).

[25] Yi HY, Zhang BJ, Gu MF, Ma HL, Zhang XH, Chen X, Xie C, Gao ZL, Shan SJ, Zhu KD, Hao Y. 2024. Discovery of isolated shoals in the Permian Maokou Formation of eastern Sichuan Basin and their natural gas exploration potential. Natural Gas Industry, 44(6), 1-11 doi: 10.3787/j.issn.1000-0976.2024.06.001. (in Chinese with English abstract).

[26] Zhang H, Wang DN, Li HX, Huang GN, Chen X. 2017. High accurate seismic data reconstruction based on non-uniform curvelet transform. Chinese Journal of Geophysics, 60(11), 4480-4490. doi: 10.6038/cjg20171132.

[27] Zhang H, Zhang HQ, Zhang JH, Hao YJ, Wang BF. 2020. An anti-aliasing POCS interpolation method for regularly undersampled seismic data using curvelet transform. Journal of Applied Geophysics, 172, 103894. doi: 10.1016/j.jappgeo.2019.103894.

[28] Zhang Y, Cao QG, Luo KP, Li LL, Liu JL. 2022. Reservoir exploration of the Permian Maokou Formation in the Sichuan Basin and enlightenment obtained. Oil and Gas Geology, 43(3), 610-620 doi: 10.11743/ogg20220310. (in Chinese with English abstract).

[29] Zhao H, Liu JW, Zhang H, Lv L, Zhang JW, Zhang GR, Li MY. 2023. Analysis of seismic characteristics and a prediction method for the Maokou Formation reservoir in Northeast Sichuan. Geophysical Prospecting for Petroleum, 62(5), 925-939 doi: 10.12431/issn.1000-1441.2023.62.05.011. (in Chinese with English abstract).

[30] Zhu DC, Zou HY, Yang ML, Li T, Li T, Li HP, Xu L, Wen L, Zhou G. 2021. The Middle Permian Maokou reservoir (southern Sichuan Basin, China): Controls on karst development and distribution. Journal of Petroleum Science and Engineering, 204, 108686. doi: 10.1016/j.petrol.2021.108686.