Abstract:Single parameter prediction usually brings poor accuracy in the sand body prediction that is achieved by using seismic attributive parameters.Nevertheless,blind using the multiple parameters as the input of neural network will make nonconvergent the learning course of the neural network.After theoretical model research,19 seismic attributive parameters in time and frequency domains can be derived from the seismic informations of objective interval so as to cope with above problems.Then 8 parameters which are very related to sand body thickness are used to predict sand bodies.In addition,the usual prediction methods are analysed and compared in detail.The trial results draw the conclusion:multiparameter neural network prediction brings satisfactory accuracy,dominant-frequency linear prediction and amplitude-frequency inversion prediction give less satisfactory accuracy,and amplitude linear prediction causes poor accuracy.When the objective interval is thicker than λ⁄4,sand thickness prediction using frequency-domain parameters is more accurate than sand thickness prediction that introduces time-domain parameters.
黄真萍, 王晓华, 王云专. 地震信息的属性参数提取和砂体预测方法[J]. 石油地球物理勘探, 1997, 32(5): 669-682.
Huang Zhenping, Wang Xiaohua, Wang Yunzkuan. Method for making both seismic attributive parameter extraction and sand body prediction. OGP, 1997, 32(5): 669-682.