Prediction Technique Research On Blast Furnace Gas In Iron And Steel Enterprise

Iron and steel industry is the pillar industry of the national economy,accounting for 26% of the total national economy.At the same time,the steel industry also consumes large energy in the national.In view of the problems of huge energy consumption of iron and steel corporation,Chinese government clearly put forward the development strategy of energy conservation and emissions reduction of iron and steel enterprises in the 12 th five-year plan.Blast furnace gas is a kind of valuable energy and reasonably utilized blast furnace gas can not only reduce the rate of consumption,Also can reduce the metallurgical energy.But the metallurgical enterprises' production automation level is low in our country,melting tendency is the extensive development mode.The centralized control of gas resources ability is poorer.Scheduling also depend on some experienced scheduling person and prediction accuracy is often not guaranteed.So,according to the characteristics of gas production and consumption of iron and steel industry and establishing the forecast model is the precondition of Energy conservation and emissions reduction and improve energy utilization.This article mainly studied the blast furnace gas system.First of all,the blast furnace gas prediction results and characteristics of the gas system and gas pipes were analyzed both domestic and overseas.For the fluctuation of blast furnace production,it is difficult to accurately forecasting.Artificial neural network model combined with grey correlation analysis method was adopted to product data of blast furnace.Then,the blast furnace gas consumption of multiple links were analyzed.Respectively,sintering,coking,coal gas stove,hot rolling.By judging,and analyzing data,got the most suitable prediction model of the four parts.Respectively is: An exponential smoothing,linear regression and particle swarm optimization of least squares support vector machine(SVM)and support vector regression model.Simulation proved that the prediction effects of consumption model.An improved particle swarm optimization algorithm was proposed when predicted the hot blast stove gas consumption.This method transformed inertia weight which is so important parameters in particle swarm and using two stages of inertia weight adjustment.In the beginning and ending of algorithm,appropriate weight was given.It made particle swarm to obtain appropriate speed and algorithm can faster convergence.By comparing several error indicators,It showed the improved particle swarm algorithm to optimized least squares support vector machine(SVM)could get higher accuracy than normal support vector machine.For the big fluctuation system like hot air stove system has good prediction ability.