基于深度学习神经网络的岩性剖面预测及应用  

Prediction and application of lithologic profile based on deep learning neural network

作　　者：史航宇 文小林 丁慧霞 齐迪 韩淑华 Shi Hangyu;Wen Xiaolin;Ding Huixia;Qi Di;Han Shuhua(Chunliang Oil Production Plant of Shengli Oilfield Company,SINOPEC,Boxing 256600,China)

机构地区：[1]中石化胜利油田分公司纯梁采油厂,山东博兴256600

出　　处：《能源与环保》2023年第11期187-192,共6页CHINA ENERGY AND ENVIRONMENTAL PROTECTION

基　　金：中石化科技项目(P21098)。

摘　　要：纯化油田储层具有岩性复杂且层多而薄的特点,这一定程度上增加了对于人工解释相应的难度,通过对研究区块内岩心及测井数据的收集整理,分析各岩性的测井曲线响应特征。以岩心资料数据结论为输出值,利用参数敏感性分析优选影响岩性的相应测井特征值作为输入值,建立深度神经网络模型,通过数据清洗参数优化等过程对样本进行训练,得出一套适用于该区块岩性分类预测的模型参数,对比深度校正后的录井结论预测准确度达到80%,最后编制岩性剖面预测模块。应用预测结论对区块内单井进行分析得出相对应的目的层段岩性,采用适合的开发生产措施提质使得增产增效效果明显,文中实例井初期日增油3.6 t,含水降低25%,得出该模型对岩性纵向展布规律认识以及对油水井后期开发调整具有一定的指导意义。The reservoirs in Chunhua Oilfield have the characteristics of complex lithology,multiple layers and thin layers,which increases the difficulty of manual interpretation to a certain extent.Through collecting and sorting the core and logging data in the study block,the response characteristics of each lithology logging curves were analyzed.With the conclusion of core data as the output value,the parameter sensitivity analysis was used to select the corresponding logging characteristic values affecting the lithology as the input values,the deep neural network model was established,and a set of model parameters suitable for lithology classification prediction of this block was obtained by training the samples through the process of data cleaning parameter optimization,etc.The prediction accuracy of the logging conclusion after depth correction reached 80%.Finally,the prediction module of lithology profile was developed.Based on the prediction results,the corresponding target interval lithology was obtained by analyzing the single well in the block,and the effect of stimulation and efficiency was obvious by adopting suitable development and production measures to improve the quality.the daily oil increase was 3.6 t,and the water content was reduced by 25%.It is concluded that this model has certain guiding significance for vertical recognition of the formation and for later development and adjustment of oil and water wells.

关 键 词：测井 岩性剖面 可视化 深度学习神经网络 

分 类 号：P631[天文地球—地质矿产勘探]