论文详情
常规的地震反演方法通常假设噪声信号服从高斯分布，以此构建出的目标函数在特定地区反演效果好，但其普适性较小。假设地震噪声信号和岩石物理模型误差服从高斯分布且带有加性脉冲噪声，引入Hampel三截尾函数构造能够同时压制高斯噪声和脉冲噪声的反演目标函数，利用牛顿法求解目标函数得到储层弹性和物性参数反演的迭代公式，并通过阈值的选择适应噪声分布不同的地区。该算法利用Hampel三截尾函数构建加权矩阵,能在一定程度上自适应调节地震数据、岩石物理约束和先验信息之间的权重，消除地震和岩石物理噪声的统计性误差造成的反演不稳定性，最终得到稳定的储层弹性和物性参数反演结果。通过理论模型测试和实际资料应用显示了算法的稳定性和可靠性。
The discrepancies between geophysical measurements and forward modeling data are commonly modeled as Gaussian errors by conventional seismic inversion methods.However,misfit functions constructed by this distribution always have good inversion result in specified regions but have weak adaptability.We assume Gaussian distribution with additional impulse distribution for both seismic noise data and elastic property deviations from the rock-physics model.In this case,Hampels three-part redescending estimate function is included in the misfit function,which could perform better to suppress both Gaussian and impulse errors as a robust measure.Minimizing the misfit function classically leads to the Newton’s optimization method,with which elastic and physical properties of reservoir are obtained.This approach can be applied to the areas with different distribution of errors by choosing the degree of threshold.The relative contributions of the rock-physics constraints and prior information are adaptively adjusted,which can increase the stability of the algorithm and ensure a better inversion.Both synthetic and real seismic data examples show that the method can obtain stable and reliable inversion results.
国家自然科学基金联合基金项目(U1562218)资助。