为研究瓦斯压力变化过程中渗透率演化规律，基于煤体代表性单元体(REV)细观尺度参数与体积应变关联特征推导了4种不同形式渗透率模型。利用实验数据对模型进行了可靠性验证，分析了有效应力引起的应变和吸附应变的变化规律，并与经典渗透率模型(P-M模型、S-D模型、CONNELL模型)进行对比验证。讨论了模型参数的敏感性，明确了REV细观尺度参数与裂隙率、渗透率的相关关系，并分析了吸附应变系数影响机理，进而对国内主要矿井渗透率极值压力进行了研究。结果表明：基于REV体积应变与REV裂隙体积应变的4种渗透率模型，其REV体积应变与REV裂隙体积应变线性相关，相关系数为初始裂隙率；体积模量与吸附应变系数分别控制有效应力和吸附应变对渗透率的影响；恒定围压、恒定孔压、恒定差压3种实验条件下，相比经典渗透率模型，4种渗透率模型对实验数据都有较好的匹配效果，但双参数拟合(体积模量与吸附应变系数)相比单参数拟合(吸附应变系数)，参数的拟合误差较大，由参数敏感性分析可知，当参数变化时(< 10%)渗透率预测结果相比真实值也有较大偏差；以沁水盆地为例，计算的煤体REV细观尺度参数a₀、b₀范围分别为1.73～46.31 μm和0.06～0.49 μm；恒定围压与恒定差压条件下，吸附应变系数与差压大小、气体类型及初始裂隙率有关；国内主要矿井CO₂极值压力在0.94～5.33 MPa，CH₄极值压力在1.06～6.94 MPa。

To investigate the evolution of coal permeability during the process of gas pressure change, four different forms of permeability models were derived based on the correlation between mesoscale parameters of coal representative elementary volume (REV) and strain-related characteristics. The reliability of the models was verified using experimental data, and the variation law of strain caused by effective stress and adsorption strain was analyzed and compared with classical permeability models (P-M model, S-D model, Connell model). The sensitivity of model parameters was discussed, the relationship between REV mesoscale parameters and fracture porosity and permeability was clarified, and the influence mechanism of adsorption strain coefficient was analyzed. Furthermore, the extreme values of permeability of major coal mines in China were studied. Results show that for the four permeability models derived in this paper, the REV volume strain satisfies a linear relationship with its fracture volume strain, and the correlation coefficient is the initial fracture porosity. The bulk modulus and adsorption strain coefficient control the influence of effective stress and adsorption strain on permeability, respectively. Under three different experimental conditions of constant confining pressure, constant pore pressure, and constant differential pressure, the four permeability models have shown a better matching effect on the experimental data compared with the classical permeability models. However, compared with single-parameter fitting, the parameter error in the double-parameter fitting (bulk modulus and adsorption strain coefficient) is larger. From the sensitivity analysis of parameters, it is observed that when the parameters change within 10%, the predicted permeability deviates significantly from the true value. Therefore, for the double or multiple-parameter fitting of permeability models, the parameter fitting errors could lead to serious distortion of the predicted permeability results. Taking the experimental data of coal samples from the Qinshui Basin as an example, the mesoscopic scale parameters a₀ and b₀ of REV were calculated to range from 1.73 to 46.31 μm and from 0.06 to 0.49 μm, respectively. Under constant confining pressure and constant differential pressure conditions, the adsorption strain coefficient was related to differential pressure, gas types, and initial fracture porosity. Furthermore, the extreme values of CO₂ gas and CH₄ gas in major coal mines in China are within the ranges of 0.94−5.33 MPa and 1.06−6.94 MPa, respectively.