Operation Optimization Modeling And Differential Evolution Algorithm For Production Process Of Reheating Furnace

The operation optimization of reheating furnace production process is to optimize the temperature setting of each furnace section so as to make the slab achieve the appropriate temperature for hot rolling and at the same time reduce energy consumption and oxide loss,which can help to improve the heating quality and decrease manufacturing cost.The study on the operation optimization of the furnace process(OOFP)not only promotes the development of iron and steel production process optimization theory,but also has an important and practical significance in improving the level of production quality and energy saving.To solve the problem,an operation optimization model considering heating quality,energy consumption,and oxide loss as the objectives is developed based on the mechanism models of the furnace process,and a variant of differential evolution(DE)algorithm is proposed.Based on the combination of mechanism and data analytics,a hybrid modeling method is developed to restructure the heat transfer model to improve its predictive precision;an improved DE is proposed to solve the hybrid model-based OOFP problem.Aiming at the integration optimization of predictive control and operation optimization of reheating furnace,rolling optimization is used to dynamically adjust the furnace temperature to ensure to achieve the objectives of increasing the reheating quality,reducing the energy consumption,and decreasing the oxide loss.The main works of this thesis are summarized as follows:1)Considering the dynamic,nonlinear,and time lag features in furnace production process,a model of OOFP is established to decide the temperature of each furnace segment to minimize the deviation of slab temperatures,energy consumption,and oxide loss.In order to simplify the differential equation expression in the heat transfer mechanism model,the collocation method based on Runge-Kutta interpolation is adopted to discrete the differential equation.Consequently,the dynamic mechanism model of the furnace production process is transformed into a static nonlinear programming model.A set of verification experiment based on the actual production process data is carried out to test the heat transfer model.The results show that the heat transfer model is able to fit the requirement of the operation optimization in practice.2)Based on the structural characteristics of OOFP problem and the advantages of powerful global search ability and fast convergence speed of DE algorithm,a novel variant of DE(O-DE)is proposed to solve the OOFP problem.In O-DE,a search space contraction strategy is designed to decrease the search space and speed up convergence;a composite mutation strategy and an adaptive parameter setting strategy are presented to improve the population diversity and enhance the searching efficiency,respectively.The numerical experimental results based on benchmark test data and practical data show that O-DE is able to solve general nonlinear programming problems and the OOFP problem,and O-DE is superior relative to other competitor algorithms.3)Because the relationship between the input and output of furnace is difficult to precisely express with heat transfer mechanism model,a hybrid modeling method is presented based on mechanism and data analytics to improve the prediction accuracy of heat transfer model.Specifically,the least square support vector machine(LS-SVM)with parameters optimized by DE is employed to dynamically compensate the temperature deviation of heat transfer mechanism model.Based on real production data,a group of testing experiments is designed to test the hybrid heat transfer model,and the numerical results show that the hybrid model achieves much higher accuracy than mechanism model.4)Based on the structural characteristics of OOFP with the hybrid heat transfer model,an improved O-DE algorithm(IO-DE)is proposed.In IO-DE,a feasible region dynamic adjustment strategy is developed to accelerate the convergence speed of the algorithm,and a population size gradually reduction strategy is adopted to improve the deep search ability of the algorithm.Based on benchmark test data and practical temperature data,the results of experiments demonstrate that IO-DE can efficiently solve general nonlinear programming problems and the OOFP problem based on the hybrid model,and the proposed algorithm outperforms other competitors.5)Due to the characteristics of the two stages including predictive control and operation optimization in the production process of furnace,the integrated optimization problem is investigated.The furnace temperature obtained by operation optimization is taken as the expectation value of predictive control,and the deviation of slab temperature is applied as the feedback value.The furnace production process is rolling optimized by adjusting the input parameters of the fuel control system of furnace to improve the reheating quality of slab,reduce the energy consumption,and decrease the oxide loss in practical dynamic process.The experimental results based on the actual production process data prove that the integrated optimization can effectively and accurately optimize the furnace production process.