基于分位数回归的长短期记忆网络QRLSTM的数据回归区间预测,多输入单输出模型 （Matlab完整程序和数据）

% QRLSTM % 数据集 列为特征，行为样本数目 clc clear load Train.mat % 训练集 Train(1,:) =[]; y = Train.demand(1:2000,:); x = Train{1:2000,3:end}; clearvars -except x y % 数据归一化 [xnorm,xopt] = mapminmax(x',0,1); xnorm = mat2cell(xnorm,size(xnorm,1),ones(1,size(xnorm,2))); [ynorm,yopt] = mapminmax(y',0,1); ynorm = ynorm'; %% % LSTM 层设置，参数设置 inputSize = size(xnorm{1},1); %数据输入x的特征维度 outputSize = size(ynorm,2); %数据输出y的维度 numhidden_units1=50; %网络单元 numhidden_units2= 20; numhidden_units3=100; %% % lstm for i = 0.1:0.1:0.9 layers = [ ... sequenceInputLayer(inputSize,'name','input') %输入层设置 lstmLayer(numhidden_units1,'Outputmode','sequence','name','hidden1') dropoutLayer(0.3,'name','dropout_1') lstmLayer(numhidden_units2,'Outputmode','last','name','hidden2') dropoutLayer(0.3,'name','drdiopout_2') fullyConnectedLayer(outputSize,'name','fullconnect') % 全连接层设置（影响输出维度）（cell层出来的输出层） % quanRegressionLayer('out',i)]; % 参数设定 opts = trainingOptions('adam', ... 'MaxEpochs',10, ... 'GradientThreshold',1,... 'ExecutionEnvironment','cpu',... 'InitialLearnRate',0.001, ... 'LearnRateSchedule','piecewise', ... 'LearnRateDropPeriod',2, ... %2个epoch后学习率更新 'LearnRateDropFactor',0.5, ... 'Shuffle','once',... % 时间序列长度 'SequenceLength',1,... 'MiniBatchSize',24,... 'Verbose',0); % % 网络训练 tic net1 = trainNetwork(xnorm,ynorm,layers,opts); net2(floor(i*10)) = trainNetwork(xnorm,ynorm,layers,opts); end % 测试集 load Test.mat Test(1,:) =[]; clear x y y = Test.demand; x = Test{:,3:end}; % 归一化 [xnorm,xopt] = mapminmax(x',0,1); xnorm = mat2cell(xnorm,size(xnorm,1),ones(1,size(xnorm,2))); [ynorm,yopt] = mapminmax(y',0,1); ynorm = ynorm'; color=[111,168,86;128,199,252;112,138,248;184,84,246]/255; % 迭代计算 % 点预测 for i = 1:length(xnorm) YpredictNorm1= net1.predict(xnorm(i)); Ypredict1(i) = mapminmax('reverse',YpredictNorm1,yopt); end % 区间预测 for i = 1:length(net2) YpredictNorm2(i,:) = net2(i).predict(xnorm); Ypredict2(i,:) = mapminmax('reverse',YpredictNorm2(i,:),yopt); end %% 可视化 x =[1:168]; figure fill([x,x(end:-1:1)],[smooth(Ypredict2(1,:),1)',smooth(Ypredict2(end,end:-1:1),1)'],'r','FaceColor',[1 0.8 0.8],'EdgeColor','none') hold on plot(Ypredict1,'-','Color',color(3,:),'LineWidth',1) hold on plot(y','-','Color',color(4,:),'LineWidth',1) legend('90%置信区间','预测值','实际值') title('区间预测') xlabel('Time') ylabel('load(KW)') title('预测周负荷概率分布') time =datetime('2022-10-1'):datetime('2022-10-7'); xticks([0:24:24*7]); xticklabels(string(time)); xtickangle(45) legend('90%置信区间','预测值','实际值') picp = PICP(Ypredict2,y); pimwp = PIMWP(Ypredict2,y); % 区间覆盖率 function picp = PICP(Ypredict,Yreal) % 维度保持一致（样本个数） if size(Ypredict,1) ~= size(Yreal,1) Ypredict = Ypredict'; end RangeForm = [ Ypredict(:,1),Ypredict(:,end)]; Num = 0; for i = 1:length(Yreal) Num = Num + (Yreal(i) >= RangeForm(i,1) && Yreal(i) <= RangeForm(i,2)); end picp = Num / length(Yreal); end % 区间平均宽度百分比 function pimwp = PIMWP(Ypredict,Yreal) % 维度保持一致（样本个数） if size(Ypredict,1) ~= size(Yreal,1) Ypredict = Ypredict'; end pimwp = 1/length(Yreal) * sum( (Ypredict(:,end) - Ypredict(:,1))... ./ Yreal) * 100; end