✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，matlab项目合作可私信。
🍎个人主页：Matlab科研工作室
🍊个人信条：格物致知。
更多Matlab仿真内容点击👇
⛄ 内容介绍
协方差矩阵自适应进化策略(CMA–ES)局部搜索能力强,具备旋转不变性,采用此策略结合贪心搜索算法实现配水系统优化。
⛄ 部分代码
function Pipe=BN_DGS(Pipe,Network_Number,PipeLength,inopts,Iter)
global G_NYTP;
count=0;
PL=PipeLength/Iter;
Type='U';
for h=1:size(Pipe.UGS_Sols,1)
if Pipe.UGS_Cost(h,end)==0
count=count+1;
end
end
if count==size(Pipe.UGS_Sols,1)
disp('There is not proper solution for applying Downward Greedy Search')
return
end
Sort_Index=Sorting_BN(Pipe);
ss=1;
for cc=Sort_Index(ss,1):size(Pipe.UGS_Sols,1)
if ss > round(size(Sort_Index,1)*0.1)
break
end
if Pipe.UGS_Cost(cc,end)~=0
for k1=1:inopts.NS
if round(Pipe.UGS_Cost(cc,k1)/10^6,3)==round((inopts.BestC/10^6),3) % saving the runtime
disp('---------Best known pipe design-----------------')
disp(['Pipe size =',mat2str(Pipe.UGS_Sols(cc,(k1-1)*(PL)+1:k1*(PL)))])
disp(['Cost Pipe =',num2str(round( Pipe.UGS_Cost(cc,k1)))])
Pipe.DGS_Sols(cc,(k1-1)*(PL)+1:k1*(PL))= Pipe.UGS_Sols(cc,(k1-1)*(PL)+1:k1*(PL));
Pipe.DGS_Cost(cc,k1) = Pipe.UGS_Cost(cc,k1);
if k1==inopts.NS
Pipe.DGS_Cost(cc,inopts.NS+1) = sum(Pipe.DGS_Cost(cc,1:inopts.NS)) ;
end
continue
end
Best_Pipe=Pipe.UGS_Sols(cc,(k1-1)*(PL)+1:k1*(PL));
[Cost_All,Cost_BS,Pressure_BS,Length_Pipes] = feval(inopts.CostF, Best_Pipe);
Sum_violation_BS = 0;
for ii=1:size(Pressure_BS ,2)-4
if Pressure_BS(ii)<inopts.CV
Sum_violation_BS=Sum_violation_BS+ abs(inopts.CV-Pressure_BS(ii));
end
end
disp('-----------------------------------')
disp(['Pipe size before applying the DGS=',mat2str(Best_Pipe)])
disp(['Cost Pipe before applying the DGS=',num2str(round( Cost_BS))])
disp(['Sum violation Pressure=',num2str(Sum_violation_BS)])
Sum_violation=Sum_violation_BS;
i=1;
while (Sum_violation ==0 )
Temp_Sol=Best_Pipe;
j=1;
while j<=size(Best_Pipe,2) % the number of pipe
Temp_Sol=Best_Pipe;
if Temp_Sol(j)==126.6
Temp_Sol(j)=113;
elseif Temp_Sol(j)==144.6
Temp_Sol(j)=126.6;
elseif Temp_Sol(j)==162.8
Temp_Sol(j)=144.6;
elseif Temp_Sol(j)==180.8
Temp_Sol(j)=162.8;
elseif Temp_Sol(j)==226.2
Temp_Sol(j)=180.8;
elseif Temp_Sol(j)==285
Temp_Sol(j)=226.2;
elseif Temp_Sol(j)==361.8
Temp_Sol(j)=285;
elseif Temp_Sol(j)==452.2
Temp_Sol(j)=361.8;
elseif Temp_Sol(j)==581.8
Temp_Sol(j)=452.2;
end
Solution(i,j).Pipe = Temp_Sol; % recording the solutions
[~,Solution(i,j).CostPipe,Solution(i,j).Pressure,Length_Pipes] = feval(inopts.CostF, Temp_Sol);
Table(j,1) = Solution(i,j).CostPipe;
Solution(i,j).Delta_Cost = abs(Solution(i,j).CostPipe-Cost_BS);
Table(j,2) = Solution(i,j).Delta_Cost;
Solution(i,j).Delta_Pre = 0;
Sum_violation = 0;
for ii=1:size(Pressure_BS ,2)-4
if Solution(i,j).Pressure (ii)<inopts.CV
Sum_violation=Sum_violation+ abs(inopts.CV-Solution(i,j).Pressure (ii));
end
end
Solution(i,j).Sum_violation = Sum_violation;
Table(j,3) = Solution(i,j).Sum_violation;
for ii=1:size(Pressure_BS ,2)-4
if Solution(i,j).Pressure(ii)>inopts.CV && Pressure_BS(ii)>inopts.CV
Solution(i,j).Delta_Pre=Solution(i,j).Delta_Pre+ abs(Pressure_BS(ii)-Solution(i,j).Pressure(ii));
end
if Solution(i,j).Pressure(ii)<=inopts.CV && Pressure_BS(ii)>inopts.CV
Solution(i,j).Delta_Pre=Solution(i,j).Delta_Pre+ abs(Solution(i,j).Pressure(ii));
end
Table(j,4) = Solution(i,j).Delta_Pre;
end
if Solution(i,j).Delta_Pre==0 && Solution(i,j).Delta_Cost==0
Solution(i,j).rate_im =0;
elseif Solution(i,j).Delta_Pre==0
Solution(i,j).rate_im = 0;
else
Solution(i,j).rate_im = Solution(i,j).Delta_Cost/Solution(i,j).Delta_Pre ;
end
Table(j,5) = Solution(i,j).rate_im;
j=j+1;
end % end while
% selecting the feasible solutions
k=1;
for ii=1:size(Best_Pipe,2)
if Solution(i,ii).Sum_violation==0 && Solution(i,ii).CostPipe < Cost_BS
Feasible(k)=Solution(i,ii);
k=k+1;
end
end
%finding the best candidate
if k==1
% i=i+1;
break
end
BestSol(i)=Feasible(1);
for i1=2:k-1
if Feasible(i1).rate_im > BestSol(i).rate_im
BestSol(i)=Feasible(i1);
end
end
Best_Pipe = BestSol(i).Pipe;
Pressure_BS = BestSol(i).Pressure;
Sum_violation_BS = BestSol(i).Sum_violation;
Cost_BS = BestSol(i).CostPipe;
disp(['Best solution sum Pressure violation after applying the DGS=',num2str(BestSol(i).Sum_violation)])
disp(['The pipe cost after applying the DGS=',num2str([BestSol(i).CostPipe])]);
disp(['The pipe sizes after applying the DGS=',mat2str(Best_Pipe)])
disp(['Number of feasiable solutions=',num2str(size(Feasible,2))])
if i > 1
if round(BestSol(i).CostPipe,2)== round(BestSol(i-1).CostPipe,2)
disp('---------- Upward Greedy Search is stopped ----------')
break % there is not any improvement
end
end
i=i+1;
clear Feasible
end % end while
PipeCost = Cost_BS;
Pipe.DGS_Sols(cc,(k1-1)*(PL)+1:k1*(PL))= Best_Pipe;
Pipe.DGS_Cost(cc,k1) = PipeCost;
if k1==inopts.NS
Pipe.DGS_Cost(cc,inopts.NS+1) =sum(Pipe.DGS_Cost(cc,1:inopts.NS)) ;
end
clear Solution
clear BestSol
clear Table
end % end for k1=1:inopts.NS
end % end if
ss=ss+1;
end % end for
disp('Downward Greedy Search is finished')
[MinCost]= min(nonzeros(Pipe.DGS_Cost(:,end)));
index = find(round(Pipe.DGS_Cost(:,end))==round(MinCost));
Pipe.XminD = Pipe.DGS_Sols(index(1),:);
Pipe.FminD = MinCost;
disp(['The best design of Downward Greedy Search =',num2str(MinCost)])
disp(['Pipe sizes 1=',mat2str([Pipe.DGS_Sols(index(1),:)]),',Pipe cost=',num2str(Pipe.DGS_Cost(index(1),end))])
end
⛄ 运行结果
⛄ 参考文献
[1]张梦蓓, 乔帅. WSN中节点分布的协方差矩阵自适应优化策略[J]. 仪表技术与传感器, 2016(2):4.
