冰蓄冷空调系统负荷预测模型和系统优化控制分析

《冰蓄冷空调系统负荷预测模型和系统优化控制分析》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、浙江大学博士学位论文摘要建立了优化控制运行费用最小化的数学模型并进行了数值计算。数值计算表明，在冷负荷小于融冰供冷功率时，优化控制就是融冰优先供冷，也就是融冰单供冷，与主机优先比较可节约运行费用24％～45％。在冷负荷大于融冰供冷最大功率时，又分两种情况：总冷负荷小于融冰供冷最大功率加一半的单主机满负荷时，优化控制是保持主机供冷负荷系数为o5，通过调节融冰来改变冷负荷，此时与主机优先比较增加能耗13～20％，节约运行费达92～11％；在冷负荷大于融冰供冷最大功率加一半的单主机满负荷时，优化控制就是融冰优先，通过蓄冰槽满负荷运行，调节制冷

2、主机来调节冷负荷，此时比主机优先控制仅增加能耗o～13％，但能节约运行费0～92％。关键词：冰蓄冷人工神经网络负荷预测优化控制浙江大学博士学位论文ABSTRACTSTODYONLOADp穴EDlCTl0NMODELAND0PTiMALCONTROLOFICESTORAGESYSTEMWuJie(InstituteofRefrigerationandCryogenicEngineering，ZhejiangUniversity)ABSTRACTIcestoragesystemrevivedin1990sbecauseitcarlmovet

3、hepowerusedfromdaytonightAsthepowerpriceofdayishigherthanthatofnight，icestoragesystemcanhelptosavepowerchargeChillerpriorityisthemostcommoncontrolstrategyforexistingicestoragesystems，ButonlyoptimalcontrolcartsignificantlyreducetheoperatingcosttotheminimumTheaccuracyofthe

4、loadpredictionisakeyforoptimizingthesystemcontrolThisthesistakestheicestoragesystembuildinginConstructBankHangzhouMansionasresearchexample，mainlydealswiththeartificialneuralnetworkmodelforloadprediction，andcomparestheoperatingcostandCOPofoptimalcontrolstrategywiththatofc

5、hillerprioritystrategy．Firstly,authorsetsupaSelfOrganizationfeatureMap(SOM)modelforidenfilyingthecoolingloadApplyingSOMtoidentifycoolingloadhasnotbeenreporteduptonow．Thisthesisinitiallyrevealstherelationshipofcoolingloadofdifferentweekdays．BackPropagation圆P)algorithmisde

6、scribedinde攮ilandvariousimprovedmethodstoBPalgorithmareintroduced．Authoralsodiscusseshowtoimprovethegeneralizationofneuralnetwork．Thetwomethodsforimprovinggeneralization，”regularization”and”earlystopping”arepresentedThedatapreprocessingtechniques，whichcanimprovetheeffici

7、encyofnetworktraining，arealsodiscussed+TemperaturepredictingANNmodelforthenext24hoursisestablishedBythismethod，MeanabsoluteerrorfMAE、isreducedto0．4512。Cfrom06663"CthatiscalculatedbyimprovedASHRAEcalculationmethod．Meanrelativeerror(MRE)isreducedto136％from202％Basedonauniqu

8、edaycoolingloadpredictingANNmodel，daycoolingloadpredicting．ANNmodelforworkdayandholidayisestablishedres