Causality Model-based Optimal Scheduling Approaches For Byproduct Gas System In Steel Industry And Their Applications

Steel industry has always been the focus of national energy conservation and emission reduction owing to its high energy consumption and seriously pollution.Byproduct gas,as a category of secondary energy generated from the production process,has the advantage of reducing the consumption of primary energy because of its high calorific value.Therefore,its optimal scheduling is playing a more and more crucial role in reducing energy costs and improving efficiency.Aiming at the actual demand of the byproduct gas system,i.e.the blast furnace gas(BFG)one,the coke oven gas(COG)one and the Linz-Donawitz converter gas(LDG)one in an iron and steel enterprise in China,a series of data-driven scheduling methods for byproduct gases are proposed as follows by considering the process characteristics of the generation and consumption.Given that the BFG system has the characteristics of large generation and consumption flow and strong fluctuation,a short-term scheduling method based on the causal time delay analysis is proposed.The influence factors are reduced based on the causal probability for the continuous variables,and the corresponding time delay is calculated,according to which the training sample sets are constructed and a multi-kernel least squares support vector machine is established to predict the gas tank level in the future.Then a modified fuzzy C-means clustering method is conducted for historical solution clustering in order to obtain the center that match the target vecter,and the most reasonable scheduling solution can be adopted by considering the feasibility evaluation indicator and the load capcity of each schedulable unit.Aming at the characteristics of discrete periodicity of generation and the fluctuation of consumption in LDG system,a causal-network-based method for its short-term scheduling problem and a granular-causality-based approach for the long-term one are proposed,respectively.Both generation and consumption uncertainties are considered in the shrot-term model,and the prediction interval is calculated based on the standard error of each variable and the scheduling solution for the coming 30 minutes is provided.On the other hand,taking the long-term production process of steel-making into account,the granular-causality model is established for long-term prediction,and then the optimal set of the scheduling solutions for the future 360 minutes are given by considering the economical and safety indicators.According to the situation that multiple categories of byproduct gases are coupling in the gas consumption process,a causal-interval-reasoning-based joint scheduling method is proposed.The time delay of the influence factors are considered,and the upper and lower boundaries of each gas tank level are predicted based on the granularity partition of the training sample sets.Then a four-layer causal network is established according to the characteristic of the energy consumption of the schedulable unit,and the scheduling solutions are further optimized based on the evaluation indicators of each partial network.Experiments of the proposed methods are carried out by employing the practical data obtained from a steel enterprise.The results indicate that the proposed methods are capable of providing superior performance when facing with different scheduling requirements of the byproduct gas systems compared with the manual ones,and can effectively ensure the production safety and reduce the gas emission and the energy consumption cost.Meanwhile,the practical software system has been developed and implemented in the energy management and control center of the enterprise,which demonstrates the reliability and effectiveness of the proposed methods.Therefore,this work is of great significance in energy conservation reduction and efficiency increase for iron and steel enterprises.