多层动态前向网络在油田系统预测中的应用  

APPLICATION OF MULTI-LAYER DYNAMIC FORWARD NETWORK TO OILFIELD SYSTEM PREDICTION

作　　者：陈立志[1] 方瑾[1] 蔡庆生[1] 刘铁男[2] 

机构地区：[1]中国科学技术大学计算机科学与技术系,合肥230027 [2]大庆石油学院,安达151400

出　　处：《模式识别与人工智能》2002年第3期363-366,共4页Pattern Recognition and Artificial Intelligence

摘　　要：针对油田系统常规的建模和动态预测方法的局限性,采用多层动态前向网络作为油田系统的辨识模型,研究了多层动态前向网络预测模型与预测技术.分析了递推预报误差学习算法的不足之处,提出了改进方案,提高算法的性能,从而更新了油田系统建模和预测方法.新方案在油田开发动态指标预测应用中,取得了很好的效果,表明本文方法的有效性.To overcome the limitation of the normal modeling and dynamic prediction techniques in oilfield system, multi-layer dynamic forward network is used as the identification model of the oilfield system, and the multi-layer dynamic forward network prediction model and techniques are studied. The shortcomings of recursive prediction error learning algorithm is analyzed, and a better scheme is proposed to improve its performance. The new techniques are applied in dynamic target prediction of oilfield development, and very good results that show the validity of the new scheme are obtained.

关 键 词：油田系统 多层动态前向网络 预测 多变量 非线性系统 递推预报误差学习算法 

分 类 号：TP29-A2[自动化与计算机技术—检测技术与自动化装置]