Sustainability-Aware Short-Term Building Cooling Demand Forecasting Using Multi-Source LSTM-Based Deep Learning

Sahibzada Jawad Hadi; Waseem Ullah Khan; Bilal Ahmad; Lubna Saif; Muhammad Ismaeel Jan; Mannahil Miftah; Sobia Hadi

Authors

Sahibzada Jawad Hadi Department of Electrical Engineering, University of Engineering and Technology Peshawar, Peshawar, Pakistan.
Waseem Ullah Khan College of Computer Science and Software Engineering, Hohai University, Nanjing, China
Bilal Ahmad Department of Electrical Engineering, University of Engineering and Technology Peshawar, Peshawar, Pakistan.
Lubna Saif Department of Institute of Computing, Muhammad Nawaz sharif University of agriculture Multan, Multan Pakistan
Muhammad Ismaeel Jan Department of Electrical Engineering, University of Engineering and Technology Peshawar, Peshawar, Pakistan.
Mannahil Miftah Department of Artificial Intelligence, National University of Computer & Emerging Sciences (NUCES) - Foundation of Advancement of Science & Technology (FAST), Karachi, Pakistan
Sobia Hadi Department of Software Engineering, Islamia College University Peshawar, Peshawar Pakistan.

Keywords:

Building Cooling Demand Forecasting, Deep Learning, Long Short-Term Memory (LSTM), Weather-Assisted Energy Prediction, Carbon-Aware Building Energy Management

Abstract

Precise short-term forecasting of building cooling demand is a necessary facilitator of power-efficient operation, economic planning, and carbon-conscious decision-making in modern power systems. Nevertheless, many of the current data-driven methods are based on a few input variables, ignore forecast-based exogenous data, or are affected by information leakage in time-series learning pipelines. In order to address these drawbacks, this paper suggests a high-fidelity deep learning model to forecast short-term cooling demand by jointly combining past load patterns, dynamic electricity price signals, grid carbon intensity signals, and multi-horizon weather forecasts. The suggested framework uses a multi-layer Long Short-Term Memory (LSTM) structure that will learn intricate temporal relationships in hourly building energy data throughout a complete annual period. The preprocessing strategy embraces a leakage-free approach and includes the alignment of data with time consistency, normalization of features, and the generation of sliding-window sequences in order to guarantee realistic and reliable performance assessment. The CityLearn dataset is used to train and test the model, and this data is a high-resolution simulated urban building energy environment of 8,760 hourly observations under various seasonal and operational conditions. The results of the experiment indicate that the given approach produces very accurate cooling demand forecasts, as the coefficient of determination (R²) is 0.9823, and the mean error of absolute percentage is below 1, which is much higher than that of traditional baseline models. Additional studies prove that the combination of forecast consistent weather variables, electricity pricing signals, and carbon intensity indicators can significantly boost prediction accuracy and operational relevance. The evaluated leakage-free building energy management system is simulated, but the leakage-free learning pipeline and multi-source input design can be directly applied to real-world systems, enabling the intelligent HVAC control, demand response, and low-carbon operational practices. Altogether, this article may help to fill the gap between deep learning approaches and sustainability-conscious decision-making in the contemporary energy infrastructure. The proposed model is designed for direct multi-step (multi-horizon) prediction of demand for cooling in the form of predictions of demand for several time steps in the future using forecast-aligned input features.

Author Biography

Sahibzada Jawad Hadi, Department of Electrical Engineering, University of Engineering and Technology Peshawar, Peshawar, Pakistan.

Researcher in Artificial Intelligence and Machine Learning, specializing in computer vision, deep learning, and transformer-based architectures. Actively engaged in interdisciplinary research bridging AI theory with practical applications in robotics, autonomous systems, intelligent perception, and environmental monitoring.

References

H. Y. Cairong Song, “A novel deep-learning framework for short-term prediction of cooling load in public buildings,” J. Clean. Prod., vol. 434, p. 139796, 2024, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0959652623039549

J. W. Liang Zhang, “A review of machine learning in building load prediction,” Appl. Energy, vol. 285, p. 116452, 2021, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0306261921000209

S. L. Chujie Lu, “Automated machine learning-based framework of heating and cooling load prediction for quick residential building design,” Energy, vol. 274, p. 127334, 2023, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0360544223007284

Y. H. Xiaofei Huang, “Hybrid forecasting model of building cooling load based on EMD-LSTM-Markov algorithm,” Energy Build., vol. 321, p. 114670, 2024, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0378778824007862

W. H. Mingxuan Zou, “Deep spatio-temporal feature fusion learning for multi-step building cooling load forecasting,” Energy Build., vol. 322, p. 114735, 2024, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S037877882400851X

D. Kim, D. Lee, H. Nam, and S. K. Joo, “Short-Term Load Forecasting for Commercial Building Using Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) Network with Similar Day Selection Model,” J. Electr. Eng. Technol. 2023 186, vol. 18, no. 6, pp. 4001–4009, Sep. 2023, doi: 10.1007/S42835-023-01660-3.

A. A. A. Gassar, “Short-Term Energy Forecasting to Improve the Estimation of Demand Response Baselines in Residential Neighborhoods: Deep Learning vs. Machine Learning,” Buildings, vol. 14, no. 7, p. 2242, 2024, [Online]. Available: https://www.mdpi.com/2075-5309/14/7/2242

K. Ullah, “Short-Term Load Forecasting: A Comprehensive Review and Simulation Study With CNN-LSTM Hybrids Approach,” IEEE Access, vol. 12, pp. 111858–111881, 2024, [Online]. Available: https://ieeexplore.ieee.org/document/10630814

S. A. Bryan Lim, “Temporal Fusion Transformers for interpretable multi-horizon time series forecasting,” Int. J. Forecast., vol. 37, no. 4, pp. 1748–1764, 2021, [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0169207021000637

Z. G. Yu Chen , Chang Liu , Junping Ge , Jianfeng Wu , Xin Zhao, “Deep learning models for forecasting electricity demand in green low-carbon supply chains,” Int. J. Low-Carbon Technol., vol. 19, pp. 2375–2382, 2024, [Online]. Available: https://academic.oup.com/ijlct/article/doi/10.1093/ijlct/ctae186/7776045

A. Y. O. Mobarak Abumohsen, “Electrical Load Forecasting Using LSTM, GRU, and RNN Algorithms,” Energies, vol. 16, no. 5, p. 2283, 2023, doi: https://doi.org/10.3390/en16052283.

K. Y. Óscar Trull, “Adaptive Bi-Directional LSTM Short-Term Load Forecasting with Improved Attention Mechanisms,” Energies, vol. 15, no. 17, p. 3709, 2024, [Online]. Available: https://www.mdpi.com/1996-1073/17/15/3709

R. Y. Tariq Limouni, “Accurate one step and multistep forecasting of very short-term PV power using LSTM-TCN model,” Renew. Energy, vol. 205, pp. 1010–1024, 2023, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S096014812300143X

S. A. Reenu Batra, “Integration of LSTM networks with gradient boosting machines (GBM) for assessing heating and cooling load requirements in building energy efficiency,” Sage J., vol. 42, no. 6, 2024, [Online]. Available: https://journals.sagepub.com/doi/10.1177/01445987241268075

B. Bohara, R. I. Fernandez, V. Gollapudi, and X. Li, “Short-Term Aggregated Residential Load Forecasting using BiLSTM and CNN-BiLSTM,” 2022 Int. Conf. Innov. Intell. Informatics, Comput. Technol. 3ICT 2022, pp. 37–43, 2022, doi: 10.1109/3ICT56508.2022.9990696.

Q. C. Anping Wan, “Short-term power load forecasting for combined heat and power using CNN-LSTM enhanced by attention mechanism,” Energy, vol. 282, p. 128274, 2023.

S. L. Guanzhong Chen, “A Systematic Review of Building Energy Consumption Prediction: From Perspectives of Load Classification, Data-Driven Frameworks, and Future Directions,” Appl. Sci., vol. 15, no. 6, p. 3086, 2025, [Online]. Available: https://www.mdpi.com/2076-3417/15/6/3086

R. H. Qi Dong, “Short-Term Electricity-Load Forecasting by deep learning: A comprehensive survey,” Eng. Appl. Artif. Intell., vol. 154, p. 110980, 20251, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0952197625009807

J. W. Huiming Lu, “A multi-source transfer learning model based on LSTM and domain adaptation for building energy prediction,” Int. J. Electr. Power Energy Syst., vol. 149, p. 109024, 2023, [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0142061523000819

T. O. T. Mustapha Habib, “A hybrid machine learning approach for the load prediction in the sustainable transition of district heating networks,” Sustain. Cities Soc., vol. 99, p. 104892, 2023, [Online]. Available: https://www.sciencedirect.com/science/article/pii/S2210670723005036

X. T. Pengwei Su, “Recent Trends in Load Forecasting Technology for the Operation Optimization of Distributed Energy System,” Energies, vol. 10, no. 9, p. 1303, 2017, [Online]. Available: https://www.mdpi.com/1996-1073/10/9/1303

A. Y.-K. Qingyao Qiao, “Feature selection strategy for machine learning methods in building energy consumption prediction,” Energy Reports, vol. 8, pp. 13621–13654, 2022, [Online]. Available: https://www.sciencedirect.com/science/article/pii/S2352484722020601