大地电磁反演-精英选择改进.zip_UME_magnetotelluric_改进_电阻率_遗传算法 反演

clear clc %先正演得到场值 T=logspace(-3,4,100); rho=[150,180,130]; h=[180,190]; lambda1=[rho,h]; rhos =mt1d(lambda1); %加入噪声 rhos=rhos+0.05.*(-1+2.*rand(size(rhos))).*rhos; e=1; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 遗传参数设置 length=[30,30,30,30,30]; a=[100,100,100,100,100]; b=[200,200,200,200,200]; p=0.2; %定义前30%为精英，不参与交叉变异，值的范围在0-1之间 CrossoverProbability=0.6; %交叉概率 MutationProbability=1; %变异概率 PopulationSize=3000; %群体大小 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 遗传算法主体 PopulationLength=sum(length); %字符串长度（个体长度） PopulationParent= StartPopulation( PopulationSize,PopulationLength ); %随机产生初始群体 k=1; while(e(end)>0.03) tic TargetFunctionValue = TargetFunction( rhos,PopulationParent,length,a,b ); %计算父代群体目标函数值 FitnessValue = CalculateFitnessValue( TargetFunctionValue ); %计算父代群体的适値 [ PopulationContinue,newFitnessValue,PopulationSave ] = EliteChoice( PopulationParent,FitnessValue,p ); %精英选择 PopulationSelect = SelectAndCopy( PopulationContinue,newFitnessValue ); %选择复制 PopulationCrossover = Crossover( PopulationSelect,CrossoverProbability ); %交叉 PopulationMutate = Mutate( PopulationCrossover,MutationProbability ); %变异 PopulationChild=[PopulationSave;PopulationMutate]; %组成新子代 ChildTargetFunctionValue = TargetFunction( rhos,PopulationChild,length,a,b ); %计算子代群体的目标函数 FitnessValue = CalculateFitnessValue( ChildTargetFunctionValue ); %计算子代群体的适値 [BestIndividual,BestFitnessValue] = CalculateBest( PopulationChild,FitnessValue ); %计算最大适値及其所对应的个体 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 解码进行计算误差 lambda(k,:) = DecodeToModel( BestIndividual,length,a,b ); e(k)=F_target( rhos,lambda(k,:) ); figure(1) plot(e) title('误差变化曲线') xlabel('迭代次数') ylabel('%') %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% 进行下一次迭代 disp(k);disp(e(end)); k=k+1; PopulationParent=PopulationChild; if(k>250) break end toc end figure(2) semilogx(T,rhos,'b--','LineWidth',1.5); hold on rhos1 =mt1d(lambda1); semilogx(T,rhos1,'g--','LineWidth',1.5) index=find(e==min(e)); rhos2=mt1d(lambda(index(1),:)); semilogx(T,rhos2,'r--','LineWidth',1.5) title('MT遗传算法一维反演对比') xlabel('时间T/s') ylabel('视电阻率') legend('模型正演之后加入噪声的曲线','模型理论正演曲线','遗传算法反演模型正演曲线',0) figure(3) % 绘制模型对比阶梯图 hh=[1,h(1),h(1)+h(2),2*(h(1)+h(2))]; rhorho=[rho,rho(end)]; stairs(hh,rhorho,'b-','LineWidth',1.5); hold on h1=[1,lambda(index(1),4),lambda(index(1),4)+lambda(index(1),5),2*(lambda(index(1),4)+lambda(index(1),5))]; rho1=[lambda(index(1),1:3),lambda(index(1),3)]; stairs(h1,rho1,'r--o','LineWidth',1.5); % set(gca, 'XLim',[0 1000]);set(gca, 'YLim',[150 200]); xlabel('深度Depth/m') ylabel('视电阻率\rho_a/(\Omega·m)') title('遗传算法反演模型与实际模型对比'); legend('实际模型','反演模型',0);