A novel deep-learning framework for short-term prediction of cooling load in public buildings

Song, Cairong; Yang, Haidong; Meng, Xian-Bing; Yang, Pan; Cai, Jianyang; Bao, Hao; Xu, Kangkang

doi:10.1016/j.jclepro.2023.139796

Cited by 7 publications

(2 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Subsequent research has extensively applied machine learning-based data-driven approaches to building load prediction [19]. Common machine learning prediction methods include Support Vector Machines (SVMs) [20][21][22], Artificial Neural Networks (ANNs) [23,24], and Deep Learning (DL) methods [25][26][27]. In the field of load prediction, various data-driven methods each play a unique role, synergistically enhancing the accuracy and efficiency of predictions.…”

Section: Introductionmentioning

confidence: 99%

“…Song et al presented a WOA-BiLSTM model for predicting energy consumption in a Beijing hospital, showing enhanced accuracy and significant MAPE improvements [36]. Song et al proposed a novel deep learning-based prediction framework, aTCN-LSTM, which integrates a gate-controlled multi-head temporal convolutional network and a sparse probabilistic self-attention mechanism with a BiLSTM network, demonstrating superior cooling load forecasting accuracy and effectiveness for HVAC systems optimization, validated through a 14-month study in a 51-story hotel [26]. Research indicates that BiLSTM, owing to its enhanced capabilities in both long-term and short-term memory, achieves superior predictive accuracy in complex data prediction scenarios, such as forecasting cold load and energy consumption.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Short-Term Load Forecasting Method for Industrial Buildings Based on Signal Decomposition and Composite Prediction Model

Zhao,

Fan

2024

Sustainability

View full text Add to dashboard Cite

Accurately predicting the cold load of industrial buildings is a crucial step in establishing an energy consumption management system for industrial constructions, which plays a significant role in advancing sustainable development. However, due to diverse influencing factors and the complex nonlinear patterns exhibited by cold load data in industrial buildings, predicting these loads poses significant challenges. This study proposes a hybrid prediction approach combining the Improved Snake Optimization Algorithm (ISOA), Variational Mode Decomposition (VMD), random forest (RF), and BiLSTM-attention. Initially, the ISOA optimizes the parameters of the VMD method, obtaining the best decomposition results for cold load data. Subsequently, RF is employed to predict components with higher frequencies, while BiLSTM-attention is utilized for components with lower frequencies. The final cold load prediction results are obtained by combining these predictions. The proposed method is validated using actual cold load data from an industrial building, and experimental results demonstrate its excellent predictive performance, making it more suitable for cold load prediction in industrial constructions compared to traditional methods. By enhancing the accuracy of cold load predictions. This approach not only improves the energy efficiency of industrial buildings but also promotes the reduction in energy consumption and carbon emissions, thus contributing to the sustainable development of the industrial sector.

show abstract