基于混合遗传算法的斜测电离图参数反演  

Inversion of oblique ionograms based on hybrid genetic algorithm

作　　者：宋欢[1] 胡耀垓[1] 赵正予[1] 姜春华[1] 

机构地区：[1]武汉大学电子信息学院,武汉430072

出　　处：《地球物理学报》2014年第3期703-714,共12页Chinese Journal of Geophysics

基　　金：国家自然科学基金(41327002;41375007);湖北省自然科学基金青年杰出人才项目(2011CDA099)资助

摘　　要：斜向探测是获取电离层状态信息的重要手段之一,对斜测电离图的反演可以得到电离层的相关结构参数.遗传算法是一种有效的并得到普遍应用的反演方法,该算法的求解不依赖于初值的选择,可以有效地减少反演问题解的非唯一性,但也存在"过早收敛"和局部搜索能力差等缺陷,从而导致反演精度下降,影响反演结果的可靠性.本文提出将基于模拟退火的混合遗传算法应用到斜测电离图的参数反演中,该算法不仅把握总体能力强,且具有较强的局部搜索能力,是遗传算法和模拟退火算法的优势互补.为了验证该算法反演结果的可靠性和稳定性,首先分别采用遗传算法、模拟退火算法和混合遗传算法对合成的电离图进行反演,反演参数包括临界频率,最大电子浓度和半厚度.通过对三种算法反演结果的对比,得出混合遗传算法的反演结果最接近真实值,需要的迭代次数也远远小于其他两种算法;通过改变种群大小和总迭代次数来判断参数值的改变对三种算法反演结果的影响,得出混合遗传算法有效地降低了参数的选取对反演结果的影响.然后用这三种反演算法对实测电离图进行反演,并将它们的反演结果与斜测链路中点的实际垂测数据进行比较,结果显示混合遗传算法84.62%的反演结果可以控制在误差范围之内,高于遗传算法(76.93%)和模拟退火算法(65.38%).这些都表明了混合遗传算法的反演结果具有较强的可靠性,在反演的寻优能力和稳定性上要明显优于遗传算法和模拟退火算法,对实测电离层图的反演具有很强的借鉴意义和应用价值.Ionospheric oblique sounding is a powerful tool for acquiring ionogram information, and the inversion of oblique ionogram can get various structural parameters. Genetic algorithm （GA） is a widely used effective inversion method which can reduce the non-uniqueness of the solution without depending on the initial estimation, but the accuracy and reliability of inversion may be decreased due to premature convergence and poor local search ability of GA. So an improved hybrid genetic algorithm （HGA） based on simulated annealing algorithm （SA） is proposed and firstly applied to the inversion of oblique ionogram. This method has a strong ability both to control whole and to search local, and realizes GA＇s and SA＇s complementary advantages. To verify the reliability and stability of HGA in detecting ionosphere, GA, SA and HGA are applied to the inversion of synthetic ionogram respectively. Inversion parameters include the critical frequency, the maximum electron concentration and half-thickness. By comparing three kinds of inversion results, the HGA＇s results are the nearest to real values, and its needed iterations is also much smaller than the others＇. By changing the population size and the total iteration times, the HGA effectively reduces the influence of parameters selection on inversion results. Then these three algorithms are also applied to the inversion of measured ionograms, and the comparisons between the inversion results and the actual vertical data in the middle of links show that the HGA has 84.62% of results in error range, which is higher than GA （76.93%） and SA （65.38%）. The results show the HGA is obviously better than the GA and SA in the stability of searching optimal solution. The improved HGA has a lot of referential significance and practical value on the inversion of ionogram.

关 键 词：斜测电离图 参数反演 混合遗传算法 遗传算法 模拟退火算法 

分 类 号：P352[天文地球—空间物理学]