Prediction Model Of Blast Furnace Gas Production And Its Application

In the smelting process of iron and steel enterprises,blast furnace gas is one of the important by-products.Its recycling level and balance of supply and demand will have a direct impact on the energy consumption level and environmental pollution level of enterprises.Rational use of blast furnace gas can not only reduce the outsourcing amount of metallurgical energy,but also reduce the emission and environmental pollution,which is of great significance to energy conservation and consumption reduction of enterprises.Therefore,forecasting the amount of blast furnace gas in advance and establishing an accurate prediction model can provide basis for gas dispatching and improve the energy utilization rate.This paper focuses on the study of blast furnace gas system.Based on the analysis of the prediction results of blast furnace gas at home and abroad,the following research is carried out on the prediction of blast furnace gas production:(1)Starting with the ironmaking process of blast furnace,this paper discusses the production process of blast furnace gas,and finds out the factors that affect the amount of blast furnace gas based on the mechanism,raw material condition and operation system,which lays a foundation for the establishment of prediction model in principle.(2)According to the complexity of the production process of blast furnace gas,The grey relational analysis method is used to analyze the importance of the influencing factors,and then the reasonable input of the model is determined to improve the prediction accuracy.In this paper,,the relational degree is calculated according to the grey relational degree analysis method and sorted to obtain the five main influencing factors:hot air pressure,cold wind flow,radar attachment,Stationary blade opening and rich oxygen.At the same time,due to the influence of field equipment and working environment,the collected data often contains noise.Therefore,in order to reduce the influence of noise on the final prediction result,this paper adopts singular spectrum analysis method to reduce noise on the original signal.(3)According to the current and nonlinear characteristics of blast furnace gas,a prediction model of blast furnace gas production based on LSTM neural network is established.Because LSTM has a "memory unit",it can learn long-term dependence and has stronger adaptability in time series data analysis.The key parameters of the network,such as the number of hidden layer units,the number of iterations and the number of batch processing,were determined after several experiments,and the Dropout size was set to prevent the over-fitting of the network.Finally,the model was compared with the previous prediction model based on BP neural network,and the rationality of the model in the prediction time series was verified.(4)In order to improve the prediction accuracy,a new combined prediction model based on PSO is proposed.The depth neural network model LSTM is combined with the traditional ARIMA model,and the weight and the key parameters of neural network are optimized simultaneously by using the optimization algorithm.The model overcomes the shortcomings of a single prediction model and greatly improves the prediction accuracy and stability.Finally,the validity of the model is verified by comparative experiments.(5)According to the prediction model of blast furnace gas production mentioned above,a blast furnace gas prediction system is designed and developed based on the actual field data.This system is programmed with C#language from the aspects of system architecture,database design and software functions.