Deep belief network based ensemble approach for cooling load forecasting of air-conditioning system

Due to the high energy consumption in buildings, cooling load forecasting plays a crucial role in the planning, control and operation of heating, ventilating and air-conditioning systems. However, cooling load data series always exhibit nonlinear and dynamic features, making it very difficult to be...

Full description

Saved in:
Bibliographic Details
Published in:Energy (Oxford) Vol. 148; pp. 269 - 282
Main Author: Fu, Guoyin
Format: Journal Article
Language:English
Published: Oxford Elsevier Ltd 01.04.2018
Elsevier BV
Subjects:
Air conditioners
Air conditioning
Air-conditioning system
algorithms
Artificial intelligence
Belief networks
Buildings
case studies
China
Cooling
Cooling load prediction
Cooling loads
Cooling systems
data collection
Decomposition
Deep belief network
Deep learning
Electricity consumption
energy
Energy consumption
Ensemble technique
Feature extraction
Forecasting
heat
Machine learning
Mathematical models
model uncertainty
Outliers (statistics)
planning
prediction
Uncertainty
China
Deep learning
Ensemble technique
Air-conditioning system
Deep belief network
Cooling load prediction
ISSN:0360-5442, 1873-6785
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Due to the high energy consumption in buildings, cooling load forecasting plays a crucial role in the planning, control and operation of heating, ventilating and air-conditioning systems. However, cooling load data series always exhibit nonlinear and dynamic features, making it very difficult to be forecasted accurately. Therefore, a novel deep learning based hybrid approach is originally proposed in this paper for deterministic cooling load forecasting with high accuracy. The approach is a hybrid of empirical mode decomposition, deep belief network and ensemble technique. Empirical mode decomposition is applied to decompose the original cooling load data series into several components with better outliers and behaviors. The hidden nonlinear features and high-level invariant structures in data are effectively extracted by layer-wise pre-training based deep belief network. In addition, ensemble technique is introduced and properly designed to mitigate the impact of uncertainties, i.e., model uncertainty and data noise, on forecasting accuracy. Case studies using real cooling load data collected from Shenzhen and Hong Kong have been implemented. The numerical results demonstrate that the proposed forecasting approach exhibits competitive performance when compared to the prediction algorithms of the state of the art. It is therefore convinced that the proposed approach has a high potential for improving the operating performance in energy systems. •Deep belief network is used to extract the hidden features in cooling load data.•A hybrid forecaster based on EMD and DBN is innovatively proposed.•The diverse errors in term of model misspecification and data noise are analyzed.•Ensemble technique is used to mitigate the diverse errors.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:0360-5442
1873-6785
DOI:10.1016/j.energy.2018.01.180