基于貝葉斯優(yōu)化的BiLSTM-Adaboost熱電廠熱負荷預測研究

打開文本圖片集

關(guān)鍵詞：供熱負荷預測；BiLSTM-Adaboost神經(jīng)網(wǎng)絡(luò)預測；貝葉斯優(yōu)化算法;超參數(shù)尋優(yōu)；預測精度中圖分類號：TM621.4 文獻標志碼：A 文章編號：2095-2945（2025）20-0001-06

Abstract：Inthispaper，aBiLSTM-AdaboostpredictionmodelbasedonBayesianoptimizationisproposedfortheheatload predictionproblemofthermalpowerplants.First，theefectsofprimarynetworkheatingparametersandmeteorologicalfactorson heatloadareconsideredcomprehensively，andthePearsoncorrelationcoeficientmethodisutilizedtoscrenthemodelinput variables.Secondly，usingthefeatureextractionabiltyofbidirectionallongshort-termmemorynetwork（BiLST）fortieeries data，AdaboostalgorithmisintroducedtointegratemultipleBiLSTMmodelstoimprovetheaccuracyandrobustnessofthe prediction;finaly，Bayesianoptimizationmethodisadoptedtooptimizethehyper-parametersofthemodeltosolvetheproblem ofreducedpredictionaccuracyduetotheperceivedimpropersetings.Simulationexperimentsarecarredoutwiththeactual operationdataofathermalpowerplant in China，andtheresultsshowthattheproposedBayesianoptimizationBiLSTMAdaboost modelhas high predictionaccuracyand stabilityin heatload predictioncomparedwith other network models.

Keywords： heating loadprediction; BiLSTM-Adaboostneural networkprediction;Bayesianoptimizationalgorithm; hyperparameter optimization;prediction accuracy

準確的供熱負荷預測對保障冬季供暖、增強調(diào)峰能力、節(jié)能減排、經(jīng)濟運行具有重要意義

傳統(tǒng)供熱負荷預測對大量歷史數(shù)據(jù)進行擬合計算，然而數(shù)學方法無法應對非線性情況。（剩余7056字）