基于梯度提升决策树算法的碳酸盐岩气藏水平井分段酸压效果评价

Evaluating staged acid fracturing of horizontal wells in carbonate gas reservoirs based on GBDT algorithm

四川盆地高石梯—磨溪区块上震旦统灯影组四段碳酸盐岩气藏为低孔、低渗储层,一般采用水平井+分段酸压技术开发。但是该气藏非均质性强,气井产能差异大,对分段酸压效果评价和酸压设计提出了巨大的挑战。基于数据驱动的思想,收集研究区15口水平井共计96段的地质、工程和试气资料。其中,地质资料包括酸压段的孔隙度、渗透率、含水饱和度、Ⅰ+Ⅱ类储层占比、Ⅲ类储层占比及非储层段占比;工程资料包括酸压段的酸液强度和排量;试气资料为折算每百米酸压段的测试产量。随后以酸压段为研究对象,建立了基于梯度提升决策树算法的酸压产能模型,该模型预测精度较高,对于测试集平均相对误差仅4.46%。并基于该模型研究了地质工程参数对酸压效果的影响,明确了渗透率和Ⅰ+Ⅱ类储层占比是影响酸压效果的关键参数。最后基于建立的产能模型,结合粒子群算法,开展了酸压施工参数优化和产能预测研究。在现场应用1口井共计8段,取得了良好的增产改造效果,且模型预测产量与实际产量较为接近,平均相对误差小于10%。研究成果有望为工区的分段酸压设计提供指导,提高单井开发效果。

Carbonate gas reservoirs of Dengying 4 Member in Gaoshiti-Moxi block of Sichuan Basin character low porosity and per-meability.Both horizontal wells and staged acid fracturing function as their common development technologies.These reservoirs,however,exhibit strong heterogeneity and great variations in productivity,which faces huge challenges in fracturing evaluation and design.So,following a data-driven thought,we acquired many geological,engineering,and production testing data from 96-staged fracturing in 15 horizontal wells of the study area.Among which,geological data cover porosity,permeability,water saturation,pro-portion of both type I and typeⅡreservoirs,proportion of typeⅢreservoirs,and proportion of non-reservoirs;engineering data of acidizing-fluid intensity and discharge at the fracturing stage as well as production test data of the converted every test production of 100 m at this stage,respectively.And then,a model based on gradient boosting decision tree(GBDT)algorithm was established for predicting the productivity after acid fracturing,obtaining higher accuracy with the average relative error of 4.46%only for testing set.The impact of geological and engineering data on fracturing performance was studied in the light of this model,supposing permeabil-ity and the proportion of both type I and typeⅡreservoirs as the key influential factors on the performance.Finally,fracturing param-eters were optimized and the productivity was predicted by using the model together with particle swarm optimization(PSO).Obvious stimulation was acquired after 8-staged acid fracturing in a certain well,and the production predicted from the model is closer to the real result,with the relative error below 10%on average.Research findings are expected to provide guidance for staged acid fracturing design in the study area.