小波神经网络预测控制研究及油田生产应用  

Research on Predictive Control Based on Wavelet Neural Network and Its Application in Oilfield Production

作　　者：刘宝[1] 杨金莹[1] 吴宗德 LIU Bao;YANG Jin-ying;WU Zong-de(School of Control Science and Engineering,China University of Petroleum(East China),Qingdao 266580,China;Qingdao Topscomm Communication Co.,Ltd.,Qingdao 266109,China)

机构地区：[1]中国石油大学(华东)控制科学与工程学院,山东青岛266580,中国石油大学(华东)控制科学与工程学院,山东青岛266580 [2]青岛鼎信通讯股份有限公司,山东青岛266109

出　　处：《控制工程》2022年第10期1793-1799,共7页Control Engineering of China

基　　金：中央高校基本科研业务费专项资金资助项目(20CX05006A);国家自然科学基金资助项目(60775052);中石油重大科技项目(ZD2019-183-007)。

摘　　要：针对油田注采系统的时变性易导致传统控制算法控制效果不理想的问题,提出一种小波神经网络预测控制算法。小波神经网络预测控制算法由小波神经网络在线学习算法、小波神经网络模型的构建和小波神经网络线性预测控制算法组成,通过实时在线调整参数,克服控制过程中时变引起的模型失配。控制仿真结果表明,小波神经网络预测控制算法相对于其他控制算法,具有较好的控制性能,并能有效地改善油田注采系统的注水控制效果。Aiming at the problem that the time-varying characteristics of the oilfield injection-production system leads to the unsatisfactory control effect of traditional control algorithms, a predictive control algorithm based on wavelet neural network is proposed in this paper. The predictive control algorithm based on wavelet neural network is composed of the online learning algorithm based on wavelet neural network, the construction of wavelet neural network model and the linear predictive control algorithm based on wavelet neural network.The model mismatch caused by the time-varying characteristics in the control process is overcome by adjusting parameters online in real time. The results of control simulation experiments show that the predictive control based on wavelet neural network predictive has better control performance than other control algorithms, and can effectively improve the water injection control of the oilfield injection-production system.

关 键 词：油田采油 预测控制 小波神经网络 时变系统 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]