ga_m.m

用标准遗传算法解决电力系统最优化问题

%%%%%%%%%%%%%%%  控制参数  %%%%%%%%%%%%%%%%%%%%
DIMENSION_NUM=40;%定义维数
PIRTICLE_NUM=100;%粒子数目
TIME=50;%循环次数
CIR_TIME=200;%叠代次数
pc=0.25;
pm=0.1;
%bit_num=20;种群数量,长度
bit_length=21;

%%%%%%%%%%%%%%%  模型参数  %%%%%%%%%%%%%%%%%%%%
P_REQUIRE=8550;%总功
%功率的上下限
pmin=[40  60  80 24 26   68 110 135 135 130    94  94 125 125 125   125 125 220 220 242   242 254 254 254 254   254 254  10  10  10   20 20 20 20 18   18 20 25 25 25 ];
pmax=[80 120 190 42 42  140 300 300 300 300   375 375 500 500 500   500 500 500 500 550   550 550 550 550 550   550 550 150 150 150   70 70 70 70 60   60 60 60 60 60];

%%%%%%%%%%%%%%%  程序参数  %%%%%%%%%%%%%%%%%%%%
PENISH_COEFFICIENT=100;%罚函数基数
t=1;%罚函数系数
%%%多样性参数%%%
xmax=500;
L=2*xmax*sqrt(DIMENSION_NUM);

%%%%%%%%%%%%%%%  程序变量  %%%%%%%%%%%%%%%%%%%%
ga_result=zeros(TIME,1);%保存每次的最好值
allvaluetemp=0;%保存多次结果的值，以便计算平均值
xtemp=zeros(CIR_TIME,1);
ytemp=zeros(CIR_TIME,1);
% ga_result_x_value=zeros(CIR_TIME,TIME,PIRTICLE_NUM);
% ga_result_x=zeros(CIR_TIME,TIME,DIMENSION_NUM,PIRTICLE_NUM);
ga_result_gbest_value=zeros(CIR_TIME,TIME);
ga_result_gbest=zeros(CIR_TIME,TIME,DIMENSION_NUM);
div=zeros(CIR_TIME,TIME);

for time=1:TIME
    tic;
    %全局最优值
    gbest=zeros(bit_length,DIMENSION_NUM);
    gbestf=zeros(1,DIMENSION_NUM);
    gbest_value=1000000;%假设这个为无穷大正数
    %产生随机0-1编码
    x=rand(bit_length,DIMENSION_NUM,PIRTICLE_NUM);
    x=abs(x)>0.5;
    x_float=zeros(DIMENSION_NUM,PIRTICLE_NUM);
    %计算对应的实数值
    f_value=zeros(1,PIRTICLE_NUM);
    for i=1 : PIRTICLE_NUM
        for j=1 : DIMENSION_NUM
            x_float(j,i)=pmin(j)+binary_to_float(x(:,j,i)')*(pmax(j)-pmin(j))/(pow2(bit_length)-1);
        end;
        f_value(i)=fvalue(x_float(:,i)',t);
    end;
    
    %开始循环
    for cirtime=1 : CIR_TIME 
        %计算适应度,结果是累计概率
        fitness=zeros(PIRTICLE_NUM,1);
        fitness=fitness_cal(f_value);
        %进行个体选择并产生新的种群
        x_new=x;
        for i=1 : PIRTICLE_NUM/2
            selectnum1=rand(1);
            for j=1 : PIRTICLE_NUM
                if selectnum1<fitness(j)
                    selectnum1=j;
                    break;
                end;            
            end;
            selectnum2=rand(1);
            for j=1 : PIRTICLE_NUM
                if selectnum2<fitness(j)
                    selectnum2=j;
                    break;
                end;            
            end;
            %交叉
            x_new=xnum_exchange(x,x_new,selectnum1,selectnum2,i,pc);
            %变异
            x_new=xnum_mutation(x_new,i,pm);
        end;
        x=x_new;
        
        %计算对应的实数值
        for i=1 : PIRTICLE_NUM
            for j=1 : DIMENSION_NUM
                x_float(j,i)=pmin(j)+binary_to_float(x(:,j,i)')*(pmax(j)-pmin(j))/(pow2(bit_length)-1);
            end;
            %计算函数值
            f_value(i)=fvalue(x_float(:,i)',t);
        end;
        
        for i=1 : PIRTICLE_NUM
            if gbest_value>f_value(i)
                %cirtime
                gbest_value=f_value(i);
                gbest=x(:,:,i);
                gbestf=x_float(:,i)';
            end;
        end;
        
        divtemp=diversity(PIRTICLE_NUM,L,DIMENSION_NUM,x_float',sum(x_float')/PIRTICLE_NUM);
        div(cirtime,time)=divtemp;
        
%         ga_result_x(cirtime,time,:,:)=x_float;        
%         ga_result_x_value(cirtime,time,:)=f_value;
        ga_result_gbest(cirtime,time,:)=gbestf;
        ga_result_gbest_value(cirtime,time)=gbest_value;      
        
        ytemp(cirtime)=ytemp(cirtime)+gbest_value;
        xtemp(cirtime)=cirtime;
    end;
    ga_result(time)=gbest_value;
    toc;
end;
allvaluetemp=sum(ga_result);
ga_result_mean=allvaluetemp/TIME;
ga_result_process_x=xtemp;
ytemp=ytemp/time;
ga_result_process_y=ytemp;
%save data
savefile='ga_result.mat';
save(savefile,'ga_result','ga_result_mean','ga_result_process_y','ga_result_process_x','div','ga_result_gbest','ga_result_gbest_value');
%all the date in process
% savefile='ga_result_large.mat';
% save(savefile,'ga_result_x','ga_result_x_value');
% plot the axis
hold on;
plot(xtemp,ytemp,'r');

clear;