基于SARIMA-LSTM组合模型的油气集输系统能耗预测  

Energy Consumption Prediction of Oil-Gas Gathering and Transportation System Based on SARIMA-LSTM Combined Model

作　　者：贺思宸 陈由旺[3] 朱英如[3] 侯磊[1] 刘珈铨 满建峰 张鑫儒 HE Sichen;CHEN Youwang;ZHU Yingru;HOU Lei;LIU Jiaquan;MAN Jianfeng;ZHANG Xinru(College of Mechanical and Transportation Engineering,China University of Petroleum(Beijing);Chinese PLA Army Logistics Academy;PetroChina Planning and Engineering Institute)

机构地区：[1]中国石油大学(北京)机械与储运工程学院 [2]中国人民解放军陆军勤务学院 [3]中国石油规划总院

出　　处：《油气田地面工程》2024年第7期82-89,共8页Oil-Gas Field Surface Engineering

基　　金：中国石油天然气集团有限公司“十四五”前瞻性基础性项目“油田智能级能源管控示范技术研究”(2021DJ6703)。

摘　　要：油气集输是油田开发生产过程的重要阶段,准确预测油气集输系统能耗能够为生产调度和能源管控提供支持。为提高油气集输系统能耗预测的准确性,针对其线性和非线性特征,综合考虑数理统计和机器学习预测方法的优缺点,提出一种基于季节性差分自回归积分滑动平均(SARIMA)和长短期记忆(LSTM)神经网络的组合预测模型。根据S油田M环状掺水油气集输系统6年的运行数据,设计组合模型的网络结构,训练组合模型的网络参数。研究结果表明:与传统的SARIMA模型和LSTM神经网络相比,组合模型对三个能耗指标的预测准确性显著提高,可为企业调整生产运行方案和优化能源管控方案提供指导和数据支持。Oil and gas gathering and transportation is an important stage of oilfield development and production process.Accurate prediction of energy consumption in oil and gas gathering and transporta-tion systems can provide support for production scheduling and energy control.In order to improve the accuracy of energy consumption prediction of oil and gas gathering and transportation systems,a com-bined prediction model based on Seasonal Autoregressive Integrated Moving Average(SARIMA)and Long Short-Term Memory(LSTM)neural network is proposed according to its linear and nonlinear characteristics,taking into account the advantages and disadvantages of mathematical statistics and ma-chine learning prediction methods.Based on the 6-year operational data of the M annular water blend-ing oil and gas gathering and transportation system in S Oilfield,the network structure of the combined model is designed,and the network parameters of the combined model are trained.The results show that compared with the traditional SARIMA model and LSTM neural network,the prediction accuracy of the combined model for the three energy consumption indicators is significantly improved,which can provide guidance and data support for enterprises to adjust production operation plans and optimize energy control plans.

关 键 词：油气集输系统 能耗预测 SARIMA模型 LSTM神经网络 组合模型 

分 类 号：TE863[石油与天然气工程—油气储运工程]