基于深度时序网络的固井泵压回归预测研究及应用
Research and Application of Cementing Pump Pressure Regression Prediction Based on Depth Time Series Network
郭磊鑫 1李毛毛 1莫峦奇1
作者信息
- 1. 西南石油大学计算机与软件学院,四川成都 610000
- 折叠
摘要
在固井作业中,泵压大小代表了目前固井作业的质量和施工状态,提前预测泵压有利于及时调整固井作业参数,提高施工效率,以及预防事故的发生.文章旨在通过固井作业现场的实时数据采集,对泵压进行精准的回归预测.利用现场实地采集的特征数据,以及依据作业机理和流程构建的计算特征,构建出与时间序列紧密相关的数据集.在此基础上,进一步提炼与泵压变化相关的特征数据,搭建深度时序模型.此模型旨在学习并捕捉具有时序特性的数据与泵压之间的复杂映射关系,从而实现对泵压的有效预测.
Abstract
In cementing operations,the size of pump pressure represents the quality and construction status of the current cementing operation. Predicting pump pressure in advance is conducive to timely adjusting cementing operation parameters,improving construction efficiency,and preventing accidents. This paper aims at accurate regression prediction of pump pressure through real-time data acquisition in cementing field. By using the feature data collected on the spot and the calculation features constructed according to the operation mechanism and process,the data set closely related to the time series is constructed. On this basis,the characteristic data related to pump pressure changes are further extracted and the depth time series model is built. This model is designed to learn and capture the complex mapping relationship between data with time series characteristics and pump pressure,so as to achieve effective prediction of pump pressure.
关键词
石油固井/时序网络/回归预测/泵压预测Key words
oil cementing/time series network/regression prediction/pump pressure prediction引用本文复制引用
基金项目
西南石油大学大学生创新创业训练计划(S202210615127)
出版年
2024
NETL
NSTL
万方数据