ANNUAL CONGRESS 2025

Deep Learning-Based Building Energy Consumption Prediction Model Considering Occupant Density Using Mobile Network Big Data: A Case Study of Seoul

Not scheduled

20m

Yildiz Technical University, Istanbul

Oral Track 05 | ENVIRONMENT AND CLIMATE

Speaker

Ms Havva Elmali (Chung-Ang University)

Description

In today's world, the rapidly increasing urban population and the corresponding rise in energy demand have made optimizing energy consumption in buildings a critical issue for urban planners and policymakers. Traditional energy consumption prediction models typically focus on static building features and environmental factors, yet they may not fully capture the influence of population density in the surrounding area. However, the relationship between building design, urban form, and energy consumption is complex, with user behavior playing a decisive role in energy consumption patterns (Holden & Norland, 2005). This study aims to overcome the limitations of traditional models by incorporating user density data into energy consumption prediction models through machine learning techniques, providing more accurate predictions for various building types across Seoul.
The research analyzes hourly mobile user data from 50m x 50m grid areas across Seoul to investigate the impact of fluctuations in occupant density on energy consumption. Using data from 2013 to 2023, the study integrates real-time occupant density data alongside static building features and environmental conditions into energy prediction models to improve prediction accuracy. The research proposes an activity-based building classification, recognizing the differences in usage patterns across various building types. For instance, energy consumption in residential buildings tends to increase during nighttime hours, indicating a correlation with the nighttime population. In contrast, energy consumption in commercial buildings peaks during working hours, reflecting their dependence on daytime activities. Public buildings, on the other hand, exhibit varying occupant density profiles throughout the day. These diverse density patterns are expected to contribute to more accurate energy consumption predictions.
The dataset used in the study not only reflects changes in user density throughout the day but also enables more precise predictions of building usage types and energy consumption profiles. This approach goes beyond traditional models to achieve higher accuracy in energy consumption predictions. Notably, a similar study by Diao et al. (2017) demonstrated that incorporating occupant density data significantly improved energy prediction models. In this context, the research highlights that traditional land use classifications may lead to inaccuracies in energy prediction models due to the dynamic nature of urban activities. By offering an activity-based classification system, this research proposes a more advanced method for understanding and optimizing energy consumption in urban areas. This approach aligns with Santos et al.'s (2019) work on utilizing urban mobility data in energy prediction models.
The research methodology includes the use of Long Short-Term Memory (LSTM) networks for time-series forecasting and Cluster Analysis for activity-based building classification. Additionally, regression models will be employed for statistical analysis to isolate the impact of user density on energy consumption and understand the relationships between variables. These analyses will provide a more detailed evaluation of factors such as building age, type, and environmental conditions on energy consumption.
The findings from this study are expected to contribute to more efficient and targeted urban energy planning and policies. Accurately analyzing the impact of fluctuations in occupant density on energy consumption can provide a significant advantage in developing energy-saving strategies. Integrating dynamic human activities and user density data into energy prediction models offers an innovative approach to deep learning-based urban energy optimization. This method can enhance the efficient management of energy distribution systems and support the development of sustainable energy policies in cities.

References

1. Holden, E. & Norland, I.T., 2005. Three challenges for the compact city as a sustainable urban form: Household consumption of energy and transport in eight residential areas in the Greater Oslo Region. Urban Studies, 42(5), pp. 935-956. Available at: https://www.jstor.org/stable/43197238 [Accessed 14 Jan. 2025].

2. Liu, H., Yu, Z., Zhang, X. & Li, Y., 2023. A review of data-driven building energy prediction. Buildings, 13(2), Article 532. Available at: https://www.mdpi.com/2075-5309/13/2/532 [Accessed 13 Jan. 2025].

3. Rahman, B., Cao, J., Huang, Y. & Li, T., 2023. Integrating urban air mobility into a public transit system: A GIS-based approach to identify candidate locations for vertiports. Vehicles, 5(4), Article 97. Available at: https://www.mdpi.com/2624-8921/5/4/97 [Accessed 15 Jan. 2025].

4. Ko, Y. (2013).Urban form and residential energy use: A review of design principles and research findings.Journal of Planning Literature.
   Available at :http://jpl.sagepub.com/content/early/2013/06/27/0885412213491499 [Accessed 10 January 2025].