基于神经网络的FPSO原油来液量软测量方法  

A Soft Sensing Method for FPSO Crude Oil Incoming Liquid Quantity Based on Neural Network

作　　者：陈长钦 刘昊 余捷 仝英利 萧阳 刘雪松 Chen Changqin;Liu Hao;Yu Jie;Tong Yingli;Xiao Yang;Liu Xuesong(CNOOC Energy Development Oil Production Service Company,Tianjin 300452,China)

机构地区：[1]中海油能源发展有限公司采油服务公司,天津300452

出　　处：《机电工程技术》2024年第1期278-281,共4页Mechanical & Electrical Engineering Technology

基　　金：海油发展重大科技专项(HFKJ-ZDZX-CY-2021-03)。

摘　　要：现有FPSO原油处理工艺流程中,由于原油携带沙子较多,经常导致流量计堵塞,从而影响原油来液测量的准确性。现场操作人员无法准确确定破乳剂的添加量,只能依靠经验进行操作,通常采用过量添加破乳剂的方式来确保处理后的原油符合要求。为了解决上述问题,采用LSTM神经网络对原油来液进行软测量建模,选取工艺流程中相关参数的测量值作为关联特征,通过训练模型并将预测结果与真实值进行对比。研究结果表明,所提方法输出结果的误差较小,可以进一步优化并推广至原油处理工艺流程的实时控制中,为实现破乳剂的自动控制提供了数据基础。In the existing FPSO crude oil treatment process,the flow meter is often blocked due to the large amount of sand carried by the incoming crude oil,resulting in inaccurate measurement of the incoming crude oil.Production operators are unable to accurately determine the amount of demulsifier added and can only rely on production experience.To ensure the quality of the treated crude oil after treatment,operators usually add excessive demulsifier.To address these issues,this paper proposes the use of neural networks for soft-sensing modeling of the crude oil incoming liquid.Relevant parameters in the process flow are selected as associated features,and the model is trained and compared with real values for prediction.The results demonstrate that this method yields small errors in the output result of this method is small,and the model can be further optimized and extended for real-time control of the crude oil treatment process.This provides a data basis for the automatic control of demulsifier usage.

关 键 词：神经网络 原油来液量 建模构建 软测量 

分 类 号：TE862[石油与天然气工程—油气储运工程]