四川盆地渝西地区足201井区深层页岩气产能主控因素分析与预测

Analysis and Prediction of Main Controlling Factors for Deep Shale Gas Productivity in the Zu201 Well Area in the Western Chongqing Region of the Sichuan Basin

摘要: 渝西地区足201井区深层页岩气产能主控因素复杂、产能预测误差较大。为探讨研究区水平井产能差异的原因，本研究以四川盆地五峰组–龙马溪组龙一段页岩储层为例，通过核主成分分析完成主控因素的筛选，通过支持向量回归算法建立SVR产能预测模型，通过多元线性回归算法建立多元线性回归模型，分别对研究区页岩气产能进行预测，并优选预测精度高的模型作为研究区产能预测模型。结果表明，研究区页岩气初期产能主控因素为压裂液量、用液强度、TOC，主要以工程因素为主；后期产能主控因素为含气量、孔隙度、TOC、含水饱和度，以地质因素为主；SVR预测模型对研究区页岩气初期和后期产能预测的相对误差分别为5.49%和4.13%，R²分别为92.13%和93.46%；多元线性回归模型对研究区页岩气初期和后期产能预测的相对误差分别为21.05%和17.01%，R²分别为77.7%和87.6%，表明SVR预测模型更加适合本研究区。本研究为页岩气产能主控因素分析与预测提供了一种新方法。

Abstract: The main controlling factors of deep shale gas production capacity in the Zu201 well area of western Chongqing are complex, and the prediction error of production capacity is relatively large. To explore the reasons for the differences in horizontal well productivity in the study area, this study takes the shale reservoir of the Longyi section of the Wufeng Formation Longmaxi Formation in the Sichuan Basin as an example. The main controlling factors are screened through kernel principal component analysis, and an SVR productivity prediction model is established through support vector regression algorithm. A multiple linear regression model is established through multiple linear regression algorithm to predict the shale gas productivity in the study area, and the model with high prediction accuracy is selected as the productivity prediction model for the study area. The results indicate that the main controlling factors for the initial shale gas production capacity in the study area are fracturing fluid volume, fluid intensity TOC, mainly based on engineering factors; The main controlling factors for later production capacity are gas content and porosity TOC, Water saturation, mainly influenced by geological factors; The relative errors of the SVR prediction model for predicting the initial and later shale gas production capacity in the study area are 5.49% and 4.13%, respectively, with R² values of 92.13% and 93.46%, respectively; The relative errors of the multiple linear regression model for predicting the initial and later shale gas production capacity in the study area were 21.05% and 17.01%, respectively, with R² values of 77.7% and 87.6%, indicating that the SVR prediction model is more suitable for this study area. This study provides a new method for analyzing and predicting the main controlling factors of shale gas production capacity.