Real-time prediction and intelligent control system for tunnel ventilation based on CFD-LSTM hybrid architecture

With the continuous expansion of urban tunnel construction and the rapid increase in the number, traditional tunnel ventilation systems face issues such as high energy consumption, delayed response, and poor control accuracy. These problems not only pose safety risks but also significantly increase operation and maintenance costs. This study focuses on the algorithm model of intelligent tunnel ventilation systems, proposing a ventilation system that combines a lightweight long short-term memory (LSTM) prediction model with computational fluid dynamics (CFD). By using the lightweight LSTM to predict the pollutant dispersion trends at key nodes in the tunnel in real time and utilizing the ventilation scheme database constructed by CFD for multi-objective optimization, the system balances energy consumption and ventilation efficiency, achieving precise on-demand control of the ventilation units. This study demonstrates that in the intelligent ventilation algorithm, the model can quickly make predictions based on tunnel air data and select the best ventilation scheme according to the predicted results, thereby reducing fan energy consumption while completing the ventilation.