our method is based on non-linear convolution model and uses artificial nervous net algorithm and simulated annealing algorithm, introducing the acoustic impedance as condition of constrained solution to avoid the hypothesis of wavelet or reflection coefficient in common seismic inversion, and adopting the control of fine seismic interpretation model in extrapolation 反演方法以非线性褶积模型为基础,引入井旁声阻抗作为解的适定性的约束条件,避开一般地震反演对地震子波及反射系数的假设。反演外推采用精细地震解释模型控制。算法上,引用人工神经网络和模拟退火算法。