EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Real-Time Traffic Forecasting in Long Road Tunnels Using LSTM Deep Learning: A Case Study of the Hai Van Pass Tunnel Complex

Trung Ngoc Vo

International Journal of Scientific Research in Civil Engineering · 24.05.2026 · 10 (3) · 48-53

DOI: 10.32628/ijsrce261033 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Scientific Research in Civil Engineering, online erschienen am 24.05.2026. Verfasst von Trung Ngoc Vo. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Real-time traffic forecasting in long road tunnels is critical for proactive intelligent transportation systems (ITS) and emergency ventilation control. Long road tunnels, such as the Hai Van Pass Tunnel Complex in Vietnam, present unique operational challenges due to high environmental risks, traffic bottlenecks, and strict safety regulations (e.g., NFPA 502 [10]). Traditional statistical forecasting methods like ARIMA fail to capture the highly non-linear, temporal dependencies of complex tunnel traffic flows under varying environmental and operational conditions. This paper proposes a deep learning framework based on Long Short-Term Memory (LSTM) networks to forecast short-term traffic volume in long road tunnels. The model was trained and validated using high-resolution, multi-lane traffic flow data collected from 172 Automatic License Plate Recognition (ALPR) cameras and SCADA system logs of the Hai Van Pass Tunnel Complex spanning a multi-year period (2021–2027). The proposed LSTM network was benchmarked against Classical Autoregressive Integrated Moving Average (ARIMA) and Gated Recurrent Unit (GRU) networks. The experimental results demonstrate that the LSTM model achieves superior forecasting accuracy, yielding a Mean Absolute Error (MAE) of 4.12 vehicles/5-min, a Root Mean Squared Error (RMSE) of 5.89 vehicles/5-min, and a Coefficient of Determination (R²) of 0.945 under standard conditions, outperforming the GRU (MAE = 4.85, R² = 0.912) and ARIMA (MAE = 8.64, R² = 0.723) models. Furthermore, the model demonstrates strong robustness during peak holiday surges, providing critical operational lead-time for tunnel ventilation adjustments and traffic dispatching, thereby significantly enhancing safety and efficiency in critical road infrastructure.

Zitieren

APA

Vo, T. N. (2026). Real-Time Traffic Forecasting in Long Road Tunnels Using LSTM Deep Learning: A Case Study of the Hai Van Pass Tunnel Complex. International Journal of Scientific Research in Civil Engineering, 10(3), 48-53. https://doi.org/10.32628/ijsrce261033

Vancouver

Vo TN. Real-Time Traffic Forecasting in Long Road Tunnels Using LSTM Deep Learning: A Case Study of the Hai Van Pass Tunnel Complex. International Journal of Scientific Research in Civil Engineering. 2026;10(3):48-53. doi:10.32628/ijsrce261033

BibTeX
@article{Vo2026,
  author = {Vo, Trung Ngoc},
  title = {Real-Time Traffic Forecasting in Long Road Tunnels Using LSTM Deep Learning: A Case Study of the Hai Van Pass Tunnel Complex},
  journal = {International Journal of Scientific Research in Civil Engineering},
  year = {2026},
  volume = {10},
  number = {3},
  pages = {48--53},
  doi = {10.32628/ijsrce261033}
}
Kostenlos starten →