【摘要】 根据神经网络、非参数回归等非线性方法的基本原理 ,对储层参数平面分布进行了预测。结果表明 ,神经网络方法能够根据井点储层参数观测数据与井旁道地震属性之间的内在联系 ,通过自学习功能 ,建立起储层参数与地震属性间较为复杂的关系 ,适应性很强。但该方法对井点储层参数及地震属性的质量要求较高 ;而非参数回归方法不必事先知道储层参数和地震属性之间的关系 ,可以避免由于模型假设与实际情况的偏差而产生的错误预测。该模型包罗广 ,适用面宽 ,能够有效地反映地震属性与储层参数间较为复杂的关系 ,在油气预测以及油藏描述中将会有广阔的应用前景。更多还原

【Abstract】 Based on the basic principles of neural network and non parametric regression analysis, the platic distribution of reservoir parameters was predicted. The results show that neural network can be used to establish a complex correlative expression between reservoir parameters and seismic attributes, according to the inner relation between the reservoir paramers observed from a well location and the local seismic attribution, and by self learning function. This method is of a high adaptability. The neural network method needs the high quality reservoir parameters and seismic attributes of well location. But then, the non parametric regression analysis does not need the relation between seismic attributes and reservoir parameters, which avoides the errors caused by the great deviation between models’ supposition and the actual situation. This method covers a wide range and a vast application scope. It can be used to establish a relatively exact prediction equation and will be widely used for oil and gas prediction and the reservoir characterization.更多还原