clear
clc
data=dlmread('数据.csv');
%% 数据处理
%按时段T把每个场景拿出来
snWind=data(:,1:3:end);
snSolar=data(:,2:3:end);
snP=data(:,3:3:end);
snP(snP==-100)=1;
randSNCount=500; %要小于16^288 。解的场景数量
gaCount=20;%粒子数
%% 初始化
ga=repmat(struct(),gaCount,1);
for I=1:gaCount
    [ selectedIndex] = initGA( randSNCount );
    %     [ selectedIndex] = evolute(selectedIndex,randSNCount );
    ga(I).selectedIndex=selectedIndex;
    [ selectedWind,selectedSolar,candidateP ] = generateCandidate( snWind,snSolar,snP,randSNCount,ga(I).selectedIndex );
    [ fitness ] = fitnessFun( selectedWind,selectedSolar,candidateP,randSNCount );
    ga(I).fitness=fitness;
end

bestFitness=0;
bestIndex=0;
for LOOP=1:10
    fprintf('第%d次迭代\n',LOOP);
    fprintf('最好适度函数%f\n',bestFitness);
    for I=1:gaCount
        fitness=ga(I).fitness;
        if fitness>bestFitness
            bestFitness=fitness;
            bestIndex=ga(I).selectedIndex;
        end
    end
    
    for I=1:gaCount
        [ selectedIndex] = evolute(selectedIndex,randSNCount,snSolar );
        ga(I).selectedIndex=selectedIndex;
        [ selectedWind,selectedSolar,candidateP ] = generateCandidate( snWind,snSolar,snP,randSNCount,ga(I).selectedIndex );
        [ fitness ] = fitnessFun( selectedWind,selectedSolar,candidateP,randSNCount );
        ga(I).selectedIndex=selectedIndex;
        ga(I).fitness=fitness;
    end
end
%利用优化后的结果给出新的场景
newSnWind=zeros(randSNCount,96);
newSnSolar=zeros(randSNCount,96);
newSnP=zeros(randSNCount,96);
for I=1:randSNCount
    onesIndex=sparse(bestIndex(I,:),1:96,1,16,96);
    newSnWind(I,:)=sum(onesIndex.*snWind);
    newSnSolar(I,:)=sum(onesIndex.*snSolar);
    newSnP(I,:)=sum(onesIndex.*snP);
end