【摘要】 快速模拟退火算法中,要求扰动后新的模型参数必须在取值范围内,对此,人们通常采用两种方法:(1)反复扰动直到新的模型参数满足要求;(2)当新的模型参数超过最大值时,参数等于最大值;新的模型参数小于最小值时,参数等于最小值.研究发现,方法1在计算过程中增加了寻找新模型参数的时间开销,计算效率明显降低;方法2当目标函数在边界或边界的小邻域存在多个极小值时,在有限的时间内,其陷入局部极小的概率大大增加,优化效果明显下降.本文在前人研究的基础上提出了改进的模拟退火算法,同时利用改进的算法反演了地层泊松比、纵波速度和密度.结果表明,改进的算法加快了模拟退火算法的收敛速度,提高了模拟退火算法的效率和精度.更多还原

【Abstract】 Newly generated parameter of model must be within the solution space in the current fast simulated annealing algorithm.So,two methods are generally used:(1) searching the proper one until the new one is in the solution space;(2) substituting the maximal value for the new one if the new one is bigger than the maximal one;contrarily,substituting the minimal value for the new one if the new one is smaller than the minimal one.Method one takes so much time in computing and lowers the efficiency;in the limited time,possibility of local optimized results increases with the object function has many local optimized results at the boundary parameters or the neighborhood of parameters.Consequently,the optimized does not appear.In order to resolve this problem,we bring forward a modified algorithm and apply it to pre-stack inversion of reservoir parameters(Poisson ratio,P-wave velocity and density).The modified algorithm reaches the optimized very faster than before,and it gets more efficient and more precise results.更多还原