Research On Decision Control Of Mine Intelligent Ventilation System Based On Deep Reinforcement Learning

The mine ventilation system is an important system to ensure the safe and healthy operation of mines.Its function is to transport fresh air from the surface to each operating area underground to dilute and eliminate toxic and harmful gases and coal dust,so as to provide a good operating environment for underground operations.Most of the traditional mine ventilation methods are simple and crude,and the ventilation decision is difficult to make realtime adjustment according to the changes of the tunnel environment,which may cause local accumulation of gas and dust and increase the risk of accidents,so the development of intelligent mine ventilation becomes inevitable.Mine intelligent ventilation decision-making is the core and key of mine intelligent ventilation construction.With the continuous development of artificial intelligence technology,deep reinforcement learning methods have great progress in solving a variety of decision-making problems,which provides new ideas for solving mine intelligent ventilation decision-making problems.To address the problem that it is difficult to realize adaptive on-demand ventilation by traditional ventilation methods,this thesis,after completing the construction of the mine intelligent ventilation system architecture,investigates the mine intelligent ventilation decision control based on deep reinforcement learning algorithm,and conducts simulation experiments and analysis,which provides some reference and support for the mine intelligent ventilation decision control.The main work of this paper is as follows:(1)To meet the demand for intelligent mine ventilation,this paper establishes the overall architecture of intelligent mine ventilation system by combining the air flow control device of air duct outlet developed by the group with the digital twin theory model.On this basis,the functional principles of each component of the system are explained,and the operation mechanism of the intelligent mine ventilation system is described in detail.Finally,the framework of the deep reinforcement learning intelligent ventilation decision-making system and its training process are designed under the overall architecture,which provides support for the subsequent research on intelligent ventilation decision-making algorithms in mines.(2)To address the problem that the traditional mine ventilation fluid dynamics model is difficult to construct due to various factors,this paper adopts a data-driven modeling approach to build a mine ventilation system model.By analyzing the characteristics of mine ventilation,the data of gas and dust concentration in the tunnel and the parameter data of the duct control device are used as the input and output parameters of the mine ventilation system model,and the end-to-end mine ventilation system modeling is completed by training the designed BP neural network,and the validity of the mine ventilation system model is verified by comparison experiments.The mine ventilation reinforcement learning environment is developed,which provides interactive environment for subsequent algorithm training.(3)To address the problem that traditional mine ventilation is difficult to achieve ondemand ventilation,this paper designs a mine intelligent ventilation decision algorithm based on two deep reinforcement learning algorithms,Deep Q Network(DQN)and Deep Deterministic Policy Gradient(DDPG),respectively.Firstly,the corresponding state action space,reward function and value policy network are designed according to the structure of the two algorithms,and then the training is carried out in the built mine ventilation reinforcement learning environment.Finally,the effectiveness of the two algorithms is verified by simulation experiments,and it is concluded that the DDPG-based mine intelligent ventilation decision algorithm can reduce the gas dust concentration in the tunnel more effectively.The deep reinforcement learning algorithm is proved to be superior to mine intelligent ventilation decision-making.(4)To address the problem that it is difficult to combine the mine ventilation decision control algorithm with the actual one,this paper encapsulates the intelligent decision algorithm and completes the development of the mine intelligent ventilation digital twin prototype system based on it.The system workflow and functions are first designed from the system requirements,then the visualization interface is written by Unity framework,and the corresponding functions are developed by C# language,and finally the functions of the prototype system are well encapsulated to improve the scalability and usability of the system.