matlab电力系统仿真-matlab光伏并网-太阳辐射预测

clear all; clc; %载入辐照度数据 load irr-temp-data.mat %提取每天8:00-18:00的辐照度数据 irr_train=trainirr(33:72,:); irr_test=testirr(33:72,:); %选择输入与期望 [M_train N_train]=size(irr_train);%训练集的行与列 [M_test N_test]=size(irr_test);%测试集的行与列 % % %% 将前三天作为输入 后一天作为输出 重新构造训练集和测试集 % % %训练集输入和期望 % for i=1:(N_train-3) % % irr_train_input(:,i)=[irr_train(:,i);irr_train(:,i+1);irr_train(:,i+2)]; % irr_train_target(:,i)=irr_train(:,i+3); % end % % for i=1:(N_test-3) % irr_test_input(:,i)=[irr_test(:,i);irr_test(:,i+1);irr_test(:,i+2)]; % irr_test_target(:,i)=irr_test(:,i+3); % end %% 将前一天数据作为输入 后一天作为输出个 构造训练集和测试集 %训练集输入和期望 irr_train_input=irr_train; for i=1:(N_train-1) irr_train_target(:,i)=irr_train(:,i+1); end %测试集输入和期望 irr_test_input=irr_test; for i=1:(N_test-1) irr_test_target(:,i)=irr_test(:,i+1); end %测试集输入和期望 %% %利用小波对数据进行分解 %%分三层 [M_train_input N_train_input]=size(irr_train_input);%训练集的行与列 [M_train_target N_train_target]=size(irr_train_target);%训练目标集的行与列 [M_test_input N_test_input]=size(irr_test_input);%测试集的行与列 [M_test_target N_test_target]=size(irr_test_target);%测试目标集的行与列 n=5;%设置分解层数 A=cell(1,n+1); A_target=cell(1,n+1); B=cell(1,n+1); B_target=cell(1,n+1); %% 训练输入集进行小波分解 for i=1:N_train_input s=irr_train_input(:,i)'; [c l]=wavedec(s,n,'db3'); for j=1:n ca(j,:)=wrcoef('a',c,l,'db3',j); cd(j,:)=wrcoef('d',c,l,'db3',j); end A{1,1}(:,i)=ca(n,:)'; for j=1:n A{1,j+1}(:,i)=cd(j,:)'; end end %% 训练目标集进行小波分解 for i=1:N_train_target s1=irr_train_target(:,i)'; [c1 l1]=wavedec(s1,n,'db3'); for j=1:n ca1(j,:)=wrcoef('a',c1,l1,'db3',j); cd1(j,:)=wrcoef('d',c1,l1,'db3',j); end A_target{1,1}(:,i)=ca1(n,:)'; for j=1:n A_target{1,j+1}(:,i)=cd1(j,:)'; end end %% 测试输入集进行小波分解 for i=1:N_test_input s2=irr_test_input(:,i)'; [c2 l2]=wavedec(s2,n,'db3'); for j=1:n ca2(j,:)=wrcoef('a',c2,l2,'db3',j); cd2(j,:)=wrcoef('d',c2,l2,'db3',j); end B{1,1}(:,i)=ca2(n,:)'; for j=1:n B{1,j+1}(:,i)=cd2(j,:)'; end end %% 进行训练和预测 for i=1:(n+1) irr_train_input1=A{1,i}(:,1:400); irr_train_target1=A_target{1,i}(:,1:400); irr_test_input1=B{1,i}(:,18:27); % irr_test_target1=B{1,i}(:,501:510); %选连样本输入输出数据归一化 [inputn,inputps]=mapminmax(irr_train_input1); [outputn,outputps]=mapminmax(irr_train_target1); %% BP网络训练 % %初始化网络结构 net=newff(inputn,outputn,10); net.trainParam.epochs=100; net.trainParam.lr=0.05; net.trainParam.goal=0.000004; %网络训练 net=train(net,inputn,outputn); %% BP网络预测 %预测数据归一化 inputn_test=mapminmax('apply',irr_test_input1,inputps); %网络预测输出 an=sim(net,inputn_test); %网络输出反归一化 BPoutput=mapminmax('reverse',an,outputps); C{1,i}=BPoutput; end sum_predict=C{1,1}; for ii=1:n sum_predict=sum_predict+C{1,ii+1}; end sum_predict=sum_predict(:); irr_test_target0=irr_test_target(:,18:27); irr_test_target1=irr_test_target0(:); %% 结果分析 figure(1) plot(sum_predict,':og') hold on plot(irr_test_target1,'-*'); legend('预测输出','期望输出') title('BP网络预测输出','fontsize',12) ylabel('函数输出','fontsize',12) xlabel('样本','fontsize',12) %预测误差 error=sum_predict-irr_test_target1; figure(2) plot(error,'-*') title('BP网络预测误差','fontsize',12) ylabel('误差','fontsize',12) xlabel('样本','fontsize',12) figure(3) plot((irr_test_target1-sum_predict)./sum_predict,'-*'); title('神经网络预测误差百分比') nn=size(error,1);%预测点的个数 RMSE=sqrt(sum(error.^2)/nn) ; MSE=sum(error.^2)/nn; MAE=sum(abs(error))/nn; MAPE=sum(abs(error./sum_predict))/nn;