页岩油压裂井产量预测方法研究
Production prediction methods for fractured shale oil wells
赵庆杰 1唐宏宝 2张乾 1冯凡 3郝华松 4白石1
作者信息
- 1. 中国石油集团渤海钻探工程公司井下技术服务公司 天津 300280
- 2. 中国石油集团渤海钻探工程公司工程技术部 天津 300457
- 3. 中国石油管道局工程有限公司第四分公司 河北廊坊 065000
- 4. 中国石油集团渤海钻探工程有限公司油气井测试分公司 河北廊坊 065007
- 折叠
摘要
为了综合考虑地质因素和工程参数对页岩油压裂井产量的影响,以大港油田页岩油井为数据来源,利用Spearman相关系数和随机森林综合筛选关键特征参数,通过数据清洗、多重填补检测剔除异常值,拓展缺失数据,构建完整的压裂井产量预测数据集.基于多层感知机神经网络模型,采用网格搜索法进行基础模型的参数调优,建立了页岩油压裂井产量预测模型,训练数据集的预测平均准确度为 92.37%.经大港 10 口页岩油井的生产数据现场应用,预测产量与实际值的平均误差为7.59%,表明该产量预测模型可综合反应地质因素和工程参数对压裂井产量的影响,使预测结果与实际生产相吻合,预测精度高,满足工程需求.
Abstract
To comprehensively consider the impact of geological factors and engineering parameters on the production of fractured shale oil wells,by using the data from shale oil wells in the Dagang oilfield and by employing Spearman's correlation coefficient and random forest techniques,key characteristic parameters were meticulously selected.Through processes of data cleansing,multiple imputation to detect and remove outliers,and expanding missing data,a complete dataset for predicting the production of fractured wells was constructed.Based on a multilayer perceptron neural network model,parameter tuning of the foundational model was achieved via grid search methodology,and a model for predicting the production of the fractured shale oil wells was established.The average prediction accuracy of the training dataset reached 92.37%.When applied to the production data from 10 shale oil wells in Dagang oilfield,the average error between predicted and actual production values was 7.59%.This indicates that the prediction model can effectively integrate geological and engineering parameters'influences on the production of fractured well,ensuring a high degree of alignment between predicted and actual production values,thereby meeting engineering requirements with high prediction accuracy.
关键词
页岩油/压裂/产量预测模型/多层感知机神经网络/现场试验/大港油田/特征参数Key words
shale oil/fracturing/production prediction model/multilayer perceptron neural network/field trial/Dagang oilfield/characteristic parameter引用本文复制引用
基金项目
中国石油集团渤海钻探工程有限公司重大技术研究项目(2023ZD02F)
出版年
2024
NETL
NSTL
万方数据