Building thermal load small sample prediction method based on simulation data and meta-learning method

The invention relates to a building thermal load small sample prediction method based on simulation data and a meta-learning method, and the method comprises the steps: determining the geometric shape of a simulation building according to the building type of a target building, setting the probabili...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: WEI LIYONG, LU DEZHI, ZHANG RONGRONG, YANG YANCHUN, LI HONGMING, DING LING, CHEN BAIXIA, GUO XIAODAN, CAI LINJING, ZHANG FAN, SHI FENG, CHEN BIN, YUAN XINRUN, LIU CHANGLI, WANG CUIMIN, ZHU BOLING, SHI LINGUANG, LI XIAOBO, SUI SHUHUI, SUN XUEWEN, HAN SHENCHAO, WANG JIAGENG, YU BO, LI MIN, WANG YINLONG, LEE GIL-JIN, CAO XIAONAN, WU MINGLEI, LIU YUDE
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention relates to a building thermal load small sample prediction method based on simulation data and a meta-learning method, and the method comprises the steps: determining the geometric shape of a simulation building according to the building type of a target building, setting the probability distribution of building parameters, and carrying out the Monte Carlo sampling, and obtaining different building parameter sets representing different use conditions of the building; performing simulation to obtain thermal load data of the building under different use conditions as source building data; constructing each group of source building data into an independent meta-task, and training on the meta-task by using a meta-learning method to obtain a thermal load small sample prediction basic model suitable for the same type of buildings; and for a target building with insufficient data, performing parameter initialization on the target task model by using the obtained parameters of the basic model, and perfo