Study On Condition Monitoring And Fault Diagnosis Of Coal Mill Based On Transfer Learning

With the progress and development of new energy power generation technology,although the market share of thermal power generation in China has declined,thermal power generation is still considered to be the most important way of power generation in China at the present stage.The fuel used in thermal power generation is coal,but because of the shortage of coal,the decline of coal quality,the rise of coal prices and other reasons,the technical requirements for thermal power generation are becoming higher and higher.The premise of ensuring the economic benefit of thermal power plant operation is to ensure the safe and stable operation of thermal power plant.Power plant pulverizing equipment is an important auxiliary equipment of thermal power plant.Its condition monitoring and fault diagnosis are of great significance for ensuring the production safety of power plant,reducing the production and operation cost of thermal power generation and improving the economic benefit of power plant.However,it is difficult to obtain all kinds of fault data of coal mill in field operation.This paper analyzes and summarizes the relevant research on modeling methods of coal mills in recent years.A representative MPS medium speed coal mill is selected to study its basic structure and working principle.A coal mill model is established based on the mass energy balance and thermodynamic relationship,and the model parameters are identified with the operation data of power plants.By summarizing the type and reason of the coal mill fault,the fault simulation is carried out to observe the change of the relevant parameters.On the basis of modeling,starting from the principle of fault occurrence,the simulation and simulation of coal mill breaking fault,full coal fault and spontaneous combustion fault are mainly carried out.The parameters obtained from simulation are used as typical samples for fault diagnosis and fault classification.The fault simulation results show that there are sufficient characteristic information in the operating parameters of the coal mill.A fault diagnosis method based on similarity comparison between the fault samples obtained from simulation and the data to be diagnosed is proposed.D-S evidence theory is used to fuse the diagnosis results on the basis of multiple evidences.Compared with the judgment method using a single diagnostic parameter,this method is more reliable.The feasibility of this method is verified by the data of coal breaking in Zhuozhou Power Plant.Then,LSTM combined with the transfer learning method is used to study the fault diagnosis between different working conditions,which solves the problem of fault diagnosis under the condition of lack of fault samples.Moreover,the transfer method has universality and can be extended to the fault diagnosis between different types of coal mills.The research method in this paper is a general method and can also be used for the state diagnosis of other equipment.