摘要
提出一种基于ANN和GA融合的自学习自适应的模糊规则提取算法,用来对油层进行识别。其方法是:首先运用人工神经网络(ANN)对训练样本进行有导师学习,网络的输入是测井属性,输出表达为网络权值和输入的函数Ψk=f(xi(WG1)ij,(WG2)jk)(其中:Ψk代表含油性类别Ck的判别函数;C1为干层;C2为水层;C3为差油层;C4为油层)。然后,以Ψk作为遗传算法(GA)中的适应度函数提取对应于类别Ck的模糊规则。最后,通过某油田oilsk81和oilsk83油井的实证研究表明,该方法能够有效地识别储层的含油性。
This paper proposed a self-adapting algorithm (ANN-GA-Cascades) for extracting fuzzy rule, which is based on fusion of ANN and GA and used for recognizing oil-bearing characteristics in reservoir. Firstly, supervised learning of training sample is performed by using neural networks, with the inputs being hte simples well-logging attribute set, and the outputs being corresponding oil-bearing characteristics Ck(C1 denotes dry layer, C2 denotes water layer, C3 denotes inferior oil layer and C4 denotes oil layer). When the prescision or the maximum iteration step is obtained, the k th output node will be the corre- sponding oil-bearing characteristics, with output function ψk = f(xi (WG 1)ij, (WG2)jk), in which ψk being the distinguish function for recognizing oil-bearing characteristics, in which (WG1)ij being the connection weight of input layer to hidder layer and in which (WG2)jk being the connection weight of hidden layer to output layer. Then use genetic algorithm (GA) to randomly select the input partition and use ψk as fitness function. In this way, he optimal chromosome will be the fuzzy rule of oil-bearing characteristics Ck. Finally, the case study of this algorithm on oil well oilsk81 and oilsk83 of Jianghan oilfield has proved to be satisfying.
出处
《系统管理学报》
北大核心
2008年第6期711-716,共6页
Journal of Systems & Management
基金
国家自然科学基金资助项目(70573101)
高等学校博士学科点专项科研基金资助项目(20070491011)
湖北省教育厅人文社会科学研究资助项目
关键词
人工神经网络
遗传算法
模糊规则
储层识别
测井属性
artificial neural networks
genetic algorithm
fuzzy rules
reservoir
well logging