Intelligent ventilation-on-demand control system for the construction of underground tunnel complex

Abstract

Traditional ventilation methods consume excessive energy but still fail to meet requirements in underground tunnel group construction. Thus, a closed-loop intelligent control system for ventilation-on-demand (VOD) was developed. To address dynamic changes in ventilation load and reduce energy consumption, firstly, the developed system calculates the real-time ventilation load and establishes a ventilation-network-based control mode to represent the ventilation system structure. The deep deterministic policy gradient (DDPG) method was then employed for the closed-loop control ensuring the required air volume in each branch of tunnel groups while minimizing energy consumption. After that, the developed closed-loop intelligent ventilation control system encompasses comprehensive perception, real analysis, real-time control, and continuous optimization. This system treats decision-making, control, and feedback as subsystems that reflect the adaptability between ventilation efficiency, construction progress, and power consumption. Finally, the end-edge-cloud-based software of the system was developed to enable remote control and display on large screens, personal computers (PCs), and mobile applications (Apps) to ensure precise and timely operation. The system was employed in tunnel group under construction at the Xulong Hydropower Station in Southwestern China, and the obtained results validate its advanced closed-loop control based on reinforcement learning (RL) and confirm its feasibility in engineering practice.

Keywords

• ventilation-on-demand
• closed-loop control
• reinforcement learning
• intelligent ventilation system
• large underground tunnels