粒子滤波非线性AVO反演方法

Nonlinear AVO inversion using particle filter

提出一种新的AVO非线性反演方法,即利用粒子滤波器来求解AVO非线性贝叶斯反演问题。利用随机带权样本逼近后验概率.论文首先论述了粒子滤波器的基本原理,包括状态转移模型与观测模型,权值预测与更新,重要性密度采样等粒子滤波器应用中的关键技术.然后建立了适合于AVO反演的粒子滤波器状态转移模型和观测模型.最后,利用该方法分别对模型数据和实际资料进行了反演计算.反演结果表明,该方法具有较好的稳定性,在AVO反演中具有的一定的应用潜力.同时对地球物理反演的其他问题求解也提供了一条新的途径.

This paper presented a new method for AVO nonlinear inversion that resolves AVO Bayesian nonlinear inversion using particle filter （PF）. The key idea is to represent the required posterior density function by a set of random samples with associated weights and to compute estimates based on these samples and weights. First the basic principle of PF is dissertated, include state transmission model and observation model, weights updating and prediction, importance density sampling, etc. Then the state transmission and observation model are established for AVO inversion. Finally, both synthetic and real data are inverted using the PF. The inversion results show that the proposed method is robust and has wide application potential for AVO inversion. Meanwhile it can provide a new approach to solve the other inversion problem of geophysics.