鄂尔多斯盆地DJ区块煤层压裂主控因素及最优区间

鄂尔多斯盆地DJ区块煤层气储量大,但受地质及工程条件的影响,区块单井产气量存在明显的差异。基于区块煤层气地质及工程生产数据,首次将随机森林(random forests, RF)与交叉验证(cross validation, CV)相结合应用于煤层气领域,并利用RF-CV从地质和工程两方面筛选了影响日均产气的主控因素。为进一步验证该方法的准确性,利用随机森林回归算法预测煤层气井的产气情况,测试井的决定系数R~2=0.85,预测效果较好并验证了RF-CV方法的准确性。基于熵值法(entropy method, EM)的逼近理想解排序法(technique for order preference similarity to ideal solution, TOPSIS)对DJ区块127口井的产气效果进行综合评价,选出相对贴近度前12的井,利用统计分析法对优选的井与全部井进行了对比分析,确定了工程主控因素的最优区间,研究结果对煤层气井的压裂规模优化和有效开发提供了借鉴。

Research on Main Controlling Factors and Optimal Interval of Coal Seam Fracturing in DJ Block of Ordos Basin

The DJ block of the Ordos Basin has large coalbed methane reserves, but due to the influence of geological and engineering conditions, there are obvious differences in the gas production of a single well in the block. Based on block coalbed methane geology and engineering production data, this paper combines Random Forests (RF) and Cross Validation (CV) in the field of coalbed methane for the first time, and uses a new method RF-CV from geology and engineering. Two aspects have screened the main controlling factors that affect the average daily gas production. In order to further verify the accuracy of the method, the random forest regression algorithm is used to predict the gas production of CBM wells. The determination coefficient of the test well is 0.85, and the prediction effect is good and verified. The accuracy of the RF-CV method is improved. The technique for order preference similarity to ideal solution (TOPSIS) based on the entropy method (Entropy Method, EM) comprehensively evaluates the gas production effect of 127 wells in the DJ block, and selects the top 12 with relative closeness. In the wells, the statistical analysis method was used to compare the optimized wells with all the wells, and the optimal interval of the main control factors of the project was determined. The research results provided a reference for the optimization of the fracturing scale and effective development of coalbed methane wells.