用改进的BP神经网络由地震属性预测测井特性

Use of Optimized BP Neural Net Work Predict log Properties from Seismic Attributes

由地震数据预测测井特性能实现测井曲线的横向外推。分析的数据由一系列与某个地震数据体相连的井的目标测井曲线组成,它利用地震数据体计算一系列基于样点的属性,导出属性的某个子集与目标测井值之间的线性或非线性变换,利用建立的统计关系,由地震属性来预测测井信息。与传统反演方法相比,该方法应用更方便并可以很大地提高分辨率。介绍了一种用改进的多层前馈神经网络由地震属性进行测井特性预测的方法,并应用理论模型和实测数据对该方法进行了验证,结果令人满意。

Predicting log properties from Seismic atributes can extrapolate log craves in landscape orientation. It establishes linear or non-linear mapping relation between a subset of samplebased seismic attributes and target log curve to predict log properties. This prediction method of easier than traditional inversion method in application and improves resolution greatly. The paper introduced a method of prediction log properties with seismic attributes by optimized BP neural net works, and it had been used to compute a theory model and a real seismic date. The results were good,