地球物理测井多参数综合识别储层流体类型的新型神经网络

建立了一种基于引力的新型神经网络模型 ,引入了合力作用机制 ,解决了储层流体类型划分的判定半径问题 ,克服了以往神经网络对任何输入数据都要产生响应 ,并给出一个分类结果的缺陷。改进了传统的距离计算方法 ,克服了传统方法易受单个参数突变的影响而引起误判的缺点。利用新型神经网络方法并结合测井资料 ,对某油田的储层流体性质及类型进行了识别 ,获得了很高的识别率。与径向基神经网络的判别结果相比 ,该神经网络的识别率明显提高。

A new neural network applied to comprehensive recognition of recognition of reservoir fluid type using geophysical well logging parameters

A new neural network was developed to identify reservoir fluid types based on gravity capturing instead of distance. The resultant force competition mechanism was introduced to determine the diagnostic radii of reservoir fluid. The new neural network has no need to respond to every input data in classification application. The former algorithm for distance calculation was improved to avoid miss judgment due to the magnitude jump of a single parameter. A field case of reservoir identification using the new neural network combined with geophysical well logs was presented. The application demonstrated the advantages and the efficiency of the new neural network.