Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
An energy consumption prediction method for HVAC systems using energy storage based on time series shifting and deep learning
Graphical abstract Display Omitted
Highlights We focused on the energy consumption of energy-storage HVAC systems. The energy consumption characteristics is peak shaving and valley filling. A time series shifting method was based on the Pearson correlation coefficient. This method was combined with GRU model for energy prediction. The adaptability of this method in non-deep learning models was investigated.
Abstract The prediction of building energy consumption plays a crucial role in responding to energy demands and achieving low-carbon control through energy saving. In this study, we focused on peak shaving and valley filling in the energy consumption of office building energy-storage HVAC systems. A time series shifting method based on the Pearson correlation coefficient was combined with a gate recurrent unit (GRU) deep learning model to predict HVAC energy consumption. Additionally, we explored the adaptability of this method to non-deep learning support vector regression (SVR) and random forest (RF) models. The results demonstrated that the Kaiser-Meyer-Olkin test values of HVAC energy consumption, meteorological parameters, and indoor environmental parameters improved by 10.8%. The GRU model exhibited a 12% reduction in the root mean square error (RMSE), and the coefficient of determination (R2) was 0.850. The RMSE of the RF model decreased to 12.24 kW·h, with a 12% increase in R2 to 0.760. The SVR model achieved an RMSE of 12.59 kW·h and an R2 of 0.793. Therefore, this method can effectively improve the prediction accuracy of models by enhancing the correlation between their input and output variables.
An energy consumption prediction method for HVAC systems using energy storage based on time series shifting and deep learning
Graphical abstract Display Omitted
Highlights We focused on the energy consumption of energy-storage HVAC systems. The energy consumption characteristics is peak shaving and valley filling. A time series shifting method was based on the Pearson correlation coefficient. This method was combined with GRU model for energy prediction. The adaptability of this method in non-deep learning models was investigated.
Abstract The prediction of building energy consumption plays a crucial role in responding to energy demands and achieving low-carbon control through energy saving. In this study, we focused on peak shaving and valley filling in the energy consumption of office building energy-storage HVAC systems. A time series shifting method based on the Pearson correlation coefficient was combined with a gate recurrent unit (GRU) deep learning model to predict HVAC energy consumption. Additionally, we explored the adaptability of this method to non-deep learning support vector regression (SVR) and random forest (RF) models. The results demonstrated that the Kaiser-Meyer-Olkin test values of HVAC energy consumption, meteorological parameters, and indoor environmental parameters improved by 10.8%. The GRU model exhibited a 12% reduction in the root mean square error (RMSE), and the coefficient of determination (R2) was 0.850. The RMSE of the RF model decreased to 12.24 kW·h, with a 12% increase in R2 to 0.760. The SVR model achieved an RMSE of 12.59 kW·h and an R2 of 0.793. Therefore, this method can effectively improve the prediction accuracy of models by enhancing the correlation between their input and output variables.
An energy consumption prediction method for HVAC systems using energy storage based on time series shifting and deep learning
Liu, Huiheng (Autor:in) / Liu, Yanchen (Autor:in) / Guo, Xun (Autor:in) / Wu, Huijun (Autor:in) / Wang, Huan (Autor:in) / Liu, Yanni (Autor:in)
Energy and Buildings ; 298
01.09.2023
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch