Data Analytics Based Dynamic Energy Allocation In Iron And Steel Area

Iron and steel area(sintering,coking,ironmaking and steelmaking)covers the consumption of multiple types of energy media and accounts for a large proportion of the total energy consumption in iron and steel industry.It is critical for steel enterprises to reduce the total cost through efficient energy allocation.In the practical production,the demand of the energy will be affected by the exogenous environment.Therefore,appropriate energy allocation plan should give full consideration of the dynamic characteristics of the production environment.Based on the characteristics of energy allocation in iron and steel area including multiple processes,multiple types of energy medium and multiple dynamic factors,the data analytics based dynamic energy allocation problem is extracted and investigated.Considering the fluctuation of unit product energy ratio caused by the change of production condition,data analytics method is used to predict the amount of energy needed to complete the current production tasks based on the real-time production data and comprehensive energy configuration of each process area,then the prediction of energy demand is used to complete the energy allocation among all processes in iron and steel area.The purpose of this research is to improve energy utilization and reduce the energy cost.The main work of this thesis is summarized as follows:1)Analyze the situation of energy consumption and regeneration in iron and steel area and propose a prediction model of energy demand based on machine learning technique with consideration of the shortcomings of empirical and linear regression methods and the difficulty of establishing energy input and output mechanism models.To improve the prediction accuracy of the standard least squares support vector machine,the influence of kernel function,kernel parameters and model parameters on the prediction accuracy is studied,exponential kernel function and adaptive kernel function are constructed based on the properties of the kernel function.Real-world data in iron and steel area is used for performance demonstration.2)Based on the prediction results of energy demand,a mathematical model of dynamic energy allocation problem is established.The cost of planned energy consumption,penalty for excessive and insufficient energy supply,and extra energy purchase cost are all considered in the model.CPLEX solver is used in the numerical experiments,which verifies the effectiveness of the model,and also show that with the increase of the problem size,the solver cannot get the optimal solution within given time.3)To deal with the large scale energy allocation problem considering the stochastic energy supply and demand,real-time energy price fluctuation,an approximate dynamic programming algorithm is developed.An approximate dynamic programming based multi-stage decision-making model for energy dynamic allocation is established,where the objective is to minimize the cost of energy allocation during the planning period,and the decision variables are the amount of energy media that are allocated to each process and the amount of purchase.The myopic and value function approximation strategies of the post-decision state are proposed to solve approximate dynamic programming model.Numerical experiments are carried out to verify the effectiveness of the approximate dynamic programming algorithm for solving large-scale stochastic dynamic problems.4)A decision support system for dynamic energy allocation in iron and steel enterprises is designed and developed.The system can give the optimized dynamic energy allocation plan during the planning period by using the energy allocation algorithm developed in this thesis.