The accurate determination of methane adsorption isotherms in coals is crucial for both the evaluation of underground coalbed methane (CBM) reserves and design of development strategies for enhancing CBM recovery. However, the experimental measurement of high-pressure methane adsorption isotherms is extremely tedious and time-consuming. This paper proposed the use of an ensemble machine learning (ML) method, namely the gradient boosting decision tree (GBDT), in order to accurately estimate methane adsorption isotherms based on coal properties in the Qinshui basin, China. The GBDT method was trained to correlate the adsorption amount with coal properties (ash, fixed carbon, moisture, vitrinite, and vitrinite reflectance) and experimental conditions (pressure, equilibrium moisture, and temperature). The results show that the estimated adsorption amounts agree well with the experimental ones, which prove the accuracy and robustness of the GBDT method. A comparison of the GBDT with two commonly used ML methods, namely the artificial neural network (ANN) and support vector machine (SVM), confirms the superiority of GBDT in terms of generalization capability and robustness. Furthermore, relative importance scanning and univariate analysis based on the constructed GBDT model were conducted, which showed that the fixed carbon and ash contents are primary factors that significantly affect the adsorption isotherms for the coal samples in this study.

Heterogeneous phase composite (HPC) flooding technology that is based on branched-preformed particle gel (B-PPG) is an important technology for enhancing oil recovery in high water-cut reservoirs. In this paper, we conducted a series of visualization experiments under the condition of developed high-permeability channels after polymer flooding, with respect to well pattern densification and adjustment, and HPC flooding and its synergistic regulation. The experiments show that for polymer-flooded reservoirs, HPC flooding can significantly reduce the water cut and increase oil recovery, but that the injected HPC system mainly advances along the high-permeability channel with limited sweep expansion. Furthermore, well pattern densification and adjustment can divert the original mainstream direction, which has a positive effect on HPC flooding, and can effectively expand the sweeping range under the synergistic effect of residual polymers. Due to the synergistic effect of multiple chemical agents in the HPC system, after well pattern densification and adjustment, the production time for HPC flooding with the water cut lower than 95% was significantly prolonged. In addition, conversion schemes, in which the original production well is converted into the injection well, are better than non-conversion schemes in terms of expanding sweep efficiency and enhancing oil recovery. Therefore, for well groups with obvious high-water-consuming channels after polymer flooding, the implementation of HPC flooding can be combined with well pattern conversion and intensification in order to further improve oil displacement.

A large number of field practices show that the water flooding characteristic curve will appear up-warping phenomenon in ultra-high water cut stage, and the conventional water flooding characteristic curve is difficult to accurately characterize. In this work, a new expression of oil–water relative permeability ratio (K_ro/K_rw) and water saturation (S_w) is proposed based on the statistical analysis of experimental data of oil–water relative permeability at high displacement multiples. The new expression has a simpler form and fewer unknown parameters. The results show that the expression can accurately fit the later section of the conventional relative permeability ratio curve, and the correlation coefficient is above 0.996. On this basis, a new type of water flooding characteristic curve suitable for the whole process of water flooding reservoir development is established by combining the reservoir engineering method. Numerical simulation and field application show that the new curve has higher accuracy and wider applicability than conventional curve. The prediction error of recoverable reserves calculated by the new curve is only 0.22%, and the error of geological reserves is less than 5%. According to the comparison between the actual data and the predicted data, the actual cumulative oil production is 2.579 × 10⁶ t, the predicted by the new curve is 2.569 × 10⁶ t, the actual ultimate oil recovery is 50.235%, and the predicted is 50.04%. The predicted value is consistent with the actual one. It provides a more reliable method to accurately predict reservoir development indexes and guides the oilfield’s subsequent decision-making.

Shale reservoirs contain omnipresent nanopores. The fluid transport phenomena on the nanoscale are significantly different from that on the macroscale. The understandings of fluid transport behavior, especially multiphase flow, are still ambiguous on the nanoscale and the traditional hydrodynamic models are insufficient to describe the fluid flow in shale. In this work, we firstly use a molecular dynamics simulation to study the oil–water two-phase flow in shale inorganic quartz nanopores and investigated the unique interfacial phenomena and their influences on fluid transport in a confined nanospace. The results of the molecular simulation revealed that the water-oil-water layered structure was formed in quartz nanopores. There is no-slip boundary condition between water and quartz surface. The density dip and the extremely low apparent viscosity of the oil–water interface region were observed. The liquid–liquid slip effect happened at the oil–water interface. Based on the nano-effects obtained by the molecular simulation, two mathematical models were proposed to describe the nanoscale oil–water two-phase flow, considering both the solid–liquid and liquid–liquid interfacial phenomena, and the performances of two mathematical models were validated. This study shed light on the flow behaviors of oil and water on the nanoscale, and provides the theoretical basis for scale-upgrading, from the nanoscale to the macroscale.

The relative permeability of coal to gas and water exerts a profound influence on fluid transport in coal seams in both primary and enhanced coalbed methane (ECBM) recovery processes where multiphase flow occurs. Unsteady-state core-flooding tests interpreted by the Johnson–Bossler–Naumann (JBN) method are commonly used to obtain the relative permeability of coal. However, the JBN method fails to capture multiple gas–water–coal interaction mechanisms, which inevitably results in inaccurate estimations of relative permeability. This paper proposes an improved assisted history matching framework using the Bayesian adaptive direct search (BADS) algorithm to interpret the relative permeability of coal from unsteady-state flooding test data. The validation results show that the BADS algorithm is significantly faster than previous algorithms in terms of convergence speed. The proposed method can accurately reproduce the true relative permeability curves without a presumption of the endpoint saturations given a small end-effect number of <0.56. As a comparison, the routine JBN method produces abnormal interpretation results (with the estimated connate water saturation ≈33% higher than and the endpoint water/gas relative permeability only ≈0.02 of the true value) under comparable conditions. The proposed framework is a promising computationally effective alternative to the JBN method to accurately derive relative permeability relations for gas–water–coal systems with multiple fluid–rock interaction mechanisms.