Real-time Reliability Prediction And Fault Diagnosis Of Gas Steam Boiler Based On Deep Learning Algorithm

Compared with coal-fired boilers,gas-fired boilers have the advantages of high thermal efficiency and low air pollution,so gas-fired boilers were gradually widely used in production and life.The boiler runs in the bad working environment of high temperature and high pressure for a long time.If it can not be found and dealt with in time,it is easy to cause major safety accidents.The boiler control system is characterized by large lag,nonlinearity,random interference and time-varying parameters,and the control effect is not ideal.At present,the monitoring technology of gas-fired boiler is mostly off-line and constant control mode.However,the heat load of gas-fired boiler changes dynamically with the end demand.In view of this,a real-time reliability prediction and fault diagnosis method of Gas Steam Boiler based on depth learning algorithm was proposed in this paper.(1)According to the four common fault types of gas-fired steam boiler,eight characteristic variables were extracted,which were drum water level,outlet steam pressure,outlet steam temperature,difference between outlet steam flow rate and boiler influent flow rate,furnace pressure,inlet smoke temperature of smoke tube heat exchanger,outlet smoke temperature of smoke tube heat exchanger and outlet smoke temperature of condenser.The Aspen Plus software was used to simulate the gas steam boiler.By disturbing the feed water flow,drum pressure,fuel flow rate and air-fuel ratio of the model,the various faults were simulated and the continuous operation data of 8 characteristic variables before and after the fault occurs were collected.(2)The adaptive neural fuzzy reasoning operation system was used to establish the corresponding evaluation rules for 8 characteristic variables,to evaluate the real-time reliability of boiler system,and to predict the time series of 8 characteristic variables by combining the long and short time memory network.Real-time reliability prediction of gas steam boiler was realized.(3)The deep neural network was used to establish a fault diagnosis model based on four common faults of gas-fired steam boiler.the five main parameters of hidden layer activation function、Batch＿Size、optimizer、learning rate and dropout were adjusted and trained.the final model started to converge about 12 iterations,and the output accuracy was higher than 92 percent.Based on the simulation experiment of gas-fired steam boiler and the continuous operation data of each running state,the real-time reliability prediction and fault diagnosis of boiler system were carried out by using depth learning algorithm.The experiment proves that the proposed method has the advantages of high prediction accuracy and good diagnostic effect,and provided a feasible method for real-time reliability prediction and fault diagnosis of boiler system.