Predictive model of energy consumption for office building by using improved GWO-BP

Building energy data analysis is a major branch of smart city development research. The usual back propagation neural network model for building energy prediction has problems of unclear physical significance, poor data generalization and low fitting accuracy. Therefore, a composite prediction model...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	Energy reports Jg. 6; S. 620 - 627
Hauptverfasser:	Tian, Ying, Yu, Junqi, Zhao, Anjun
Format:	Journal Article
Sprache:	Englisch
Veröffentlicht:	Amsterdam Elsevier 01.11.2020 Elsevier Ltd
Schlagworte:	BP neural network Building electrical load prediction Fuzzy C-means clustering algorithm Grey wolf optimizer algorithm Fuzzy C-means clustering algorithm Grey wolf optimizer algorithm BP neural network Building electrical load prediction
ISSN:	2352-4847, 2352-4847
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	Building energy data analysis is a major branch of smart city development research. The usual back propagation neural network model for building energy prediction has problems of unclear physical significance, poor data generalization and low fitting accuracy. Therefore, a composite prediction model of building power consumption based on FCM-GWO-BP neural network was proposed. According to the similar statistical distribution characteristics of data, the fuzzy C-means clustering algorithm (FCM) was used to cluster the historical power consumption data. BP neural network prediction model was established for different categories to reduce the impact of relevant noise in the sample data on the modeling accuracy. Then, according to the train and test data sets of each category, the corresponding grey wolf algorithm was established to optimize the error back propagation neural network prediction model (GWO-BP). The experimental results showed that compared with the sample prediction accuracy index root mean square percentage error (RMSPE), the GWO-BP neural network after FCM clustering was reduced by about 0.225 compared with the BP model, and was reduced by about 0.135 compared with the GWO-BP model, so its prediction accuracy was improved by 75% at most. Respectively, the mean absolute percentage error (MAPE) was reduced by 14.41% and 6.48%. It can be seen that this model has strong generalization ability, better prediction accuracy and reliability, and absolutely can meet the needs of practical engineering.
ISSN:	2352-4847 2352-4847
DOI:	10.1016/j.egyr.2020.03.003