遗传算法结合正则化方法反演海洋大气波导  

Genetic algorithm with regularization method to retrieve ocean atmosphere duct

作　　者：何然[1] 黄思训[1] 周晨腾[1] 姜祝辉[1] 

机构地区：[1]解放军理工大学气象学院,南京211101

出　　处：《物理学报》2012年第4期504-511,共8页Acta Physica Sinica

基　　金：国家自然科学基金(批准号:41175025)资助的课题~~

摘　　要：针对正则化方法在解决实际反演问题时既能克服问题的不适定性又可以很大程度上抑制噪声和误差的传播,本文提出了利用遗传算法结合正则化方法的新算法,在遗传算法适应度函数中引入正则化项来反演波导参数;然后对算法进行仿真试验,结果表明新算法与传统遗传算法相比具有较高的反演精度,并指出当噪声误差小于10%时,算法具有较强的"去噪"性能;最后利用机载雷达在Wallops岛探测的海表面处局部回波资料进行反演试验,将反演结果与实测大气折射率廓线进行比较,说明该算法的有效性.新方法为海洋大气波导反演研究提供了一种新思路.Due to the fact that regularization method can overcome the ill-posed characteristic and obviously suppress noise and errors propagation in practical inverse problem,a new algorithm（genetic algorithm with regularization method） is proposed,combining fitness function of genetic algorithm with regularization term to retrieve ocean duct parameter.Simulation experiments show that retrieval accuracy of the new algorithm is better than that of traditional genetic algorithm.When the noise errors are less than 10%,the noise-immune performance is better.Finally,the inversion result from airborne radar clutter local date of sea surface in the Wallops island is compared with real refractivity profile,pointing out the efficiency of algorithm.The new method provides a new consideration for inverse research of ocean atmosphere duct.

关 键 词：遗传算法 正则化 反演 大气波导 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] P732[自动化与计算机技术—控制科学与工程]