The optimization and regulation of mine ventilation networks are essential for the safe, stable, and economical operation of these systems. The random and dynamic changes in the structure and state parameters of the ventilation network present significant challenges for optimization and decision-making. This paper addresses the stochastic uncertainties inherent in mine ventilation systems and proposes a robust optimization and control method based on deep reinforcement learning. Initially, the airflow optimization problem is mathematically formulated as a Markov decision process, eliminating the need to model and predict system uncertainties. Subsequently, an improved distributed proximal policy optimization algorithm is employed to dynamically optimize and make decisions regarding continuous airflow, directly yielding optimized control solutions for varying airflow demands. Experimental results indicate that the proposed method effectively mitigates multiple uncertainties in the ventilation network and reduces the energy consumption of mine ventilation fans.