基于GA-BP神经网络的特低渗砂岩油藏水平井产能预测——以子长油田长6油藏为例  

Prediction of horizontal well productivity in ultra-low permeability sandstone reservoirs based on GA-BP neural network:a case study of Chang 6 reservoir in Zichang Oilfield

作　　者：强璐 任宇飞 马焕焕 李刚[1] 白婷 李书琴 QIANG Lu;REN Yufei;MA Huanhuan;LI Gang;BAI Ting;LI Shuqin(Exploration&Development Technology Research Center of Yanchang Oilfield Co.,Ltd.,Yan’an 716000,Shaanxi,China)

机构地区：[1]延长油田股份有限公司勘探开发技术研究中心,陕西延安716000

出　　处：《石油地质与工程》2024年第6期63-67,共5页Petroleum Geology and Engineering

基　　金：延长油田股份有限公司科技项目“水平井高效开发配套工程技术研究与应用”(ycsy2021ky-B-15-6)。

摘　　要：为明确子长油田特低渗透砂岩油藏水平井产能主控因素,并提高BP神经网络产能预测模型准确率,运用灰色关联方法厘定影响子长油田长6油藏水平井压后产能的主控因素,并建立遗传算法优化BP神经网络的特低渗透砂岩油藏水平井产能预测模型,以子长油田长6油藏92口压裂水平井数据进行训练和验证,对比分析标准BP神经网络与GA-BP神经网络模型预测结果和误差。结果表明,地质因素和工程因素对特低渗砂岩油藏水平井产能影响程度最大,尤其是压裂段数、渗透率和孔隙度影响最为明显;GA-BP模型各项参数的预测误差均远小于标准BP模型的预测误差,平均绝对误差仅为0.76 t/d,比标准BP模型降低了75.00%;GA-BP神经网络模型具有预测结果准确度高、高效可行的特点。利用GA-BP神经网络模型预测特低渗砂岩油藏水平井产能,为制定合理开发决策提供依据,可以有效地指导现场生产。To identify the main controlling factors of the horizontal well productivity in ultra-Low permeability sandstone reservoirs in Zichang Oilfield,and to improve the accuracy of the BP neural network productivity prediction model,the gray correlation analysis was used to determine the main controlling factors affecting the horizontal well pressure productivity in Chang 6 oil reservoir of Zichang Oilfield.Based on this,a genetic algorithm was used to optimize the BP neural network for predicting the productivity.The data of 92 fractured horizontal wells in the Chang 6 oil reservoir of Zichang Oilfield were used for training and validation.The prediction results and errors of the standard BP neural network and GA-BP neural network models were compared and analyzed.The results show that geological and engineering factors have the greatest influence,especially the number of fracturing sections,permeability and porosity.The prediction errors of various parameters in the GA-BP model are much smaller than those of the standard BP model,with an average absolute error of only 0.76 t/d,which is 75.00%lower than the standard BP model.The GA-BP neural network model has the characteristics of high prediction accuracy,efficiency and feasibility.Using the GA-BP neural network model to predict the productivity provides a basis for making reasonable development decisions and effectively guide field production.

关 键 词：神经网络 特低渗油藏 产能预测 灰色关联分析法 遗传算法 

分 类 号：TE319[石油与天然气工程—油气田开发工程]