人工神经网络方法在油气层综合解释中的应用

APPLICATION OF ARTIFICIAL NEURAL NETWORKS TO INTEGRATIVE INTERPRETAT1ON O F PAY BEDS

根据油气层综合解释的特点，提出了单井油气层综合解释的人工神经网络方法。采用的神经网络模型为包含２８个输入神经元、２６个隐含神经元和４个输出神经元的３层网络。选择了松辽盆地北部大庆长垣以西地区９１个试油层段的资料对网络进行了训练，通过对已知样品的交叉验证，训练后的神经网络识别符合率为０．９３４，证明人工神经网络是一种有效的油气层综合解释方法。

Based upon the characteristics of integrative interpretation of pay beds,a new methedusing AN N for integrative pay beds interpretation is presen ted.The network consists ofthree layers,he input,hidden and output layers which have 28,26 and 4 neurons respec-tively.To try the network,91 pay beds are selected from the western area of Daqing oilfields. By overlap validation of 91 samples, 93.4 percent of the interpretations are sound. The results confirm that the new methed can be used for field interpretation.