【摘要】 微地震监测难以拾取准确初至,为了提高反演定位精度和减小多解性,研究了微地震贝叶斯差分进化反演方法.从分析讨论理论模型反演残差及其协方差分布特征出发,结合对比不加噪音和加入不同程度的噪音后残差协方差极小点位置移动、分布梯度变化特征,提出了先验信息解估计方法.针对后验估计中,由于难以获得先验信息解的方差估计致使无法计算加权系数问题,通过分析残差变化特征和解的变化关系,研究了利用残差求取加权系数的方法.为了加快寻优速度,讨论了差分进化反演方法,在变异操作方面使用差分策略,即利用种群中个体间的差分向量对个体进行扰动,实现个体变异,充分有效利用群体分布特性,提高算法的搜索能力,避免遗传算法中变异方式的不足.通过理论模型测试本方法的反演效果,并且和搜索方法反结果进行比较.测试结果证明本反演方法,对于不同程度初至干扰,反演结果向准确解逼近程度比搜索方法要好得多,实际资料的反演结果也好于搜索方法.更多还原

【Abstract】 Micro-seismic monitoring is difficult to pick up an accurate first wave.In order to improve the positioning accuracy of the inversion and reduce the problems of multiple solutions,this paper studies micro-seismic inversion method based on Bayesian theory.We discuss and analyze the statistical distribution of the theoretical model inversion residuals and covariance.We combine and compare the minimum point position and distribution gradients of residual covariance without noise and adding different levels of noise.We propose the priori information estimation method of the solution.In posteriori estimation,variance estimation of the solution space is difficult to obtain,that makes it impossible to calculate the weighting coefficients.By analyzing residual variation and the changes relationship of the solution,reconciliation of changes in the relationship by analyzing the residual variation,we try to study the method of using the residuals to obtain weighted coefficient.In order to accelerate the speed of optimization,we discuss the differential evolution inversion method.We use differential strategy in the mutation operation.The difference vector between individuals in the population disturbances individual is used to achieve individual variability.We take advantage of the mass distribution characteristics,improve the search capabilities of the algorithm,and avoid the shortcomings of variation manner in the genetic algorithm.We test the inversion effect by the theoretical model and compare the results of the search method at the same time.The test results prove that the inversion method works well.For different degrees of first arrival to interference,the approximation degree of inversion results to the exact solution is much larger than the search method.The inversion results of actual data are also better than the search method.更多还原