On Line Optimization Method Of Energy Efficiency Scheduling Of Electro-fused Magnesium Furnace Based On Demand Perdiction

The electric melting furnace is the equipment for smelting the electric melting magnesia.Its products have the properties of compact structure,strong pressure resistance,strong insulation and high heat resistance,and play an important role in metallurgy,building materials,national defense and other fields.However,it is of great significance to optimize the power supply curve and improve the energy efficiency by adopting the energy efficiency scheduling method.The traditional scheduling method of electrofused magnesium furnace,which relies on manual decision making,only takes a single electric furnace as the optimal control object and sets the electric load of a single electric furnace,but it cannot comprehensively consider the difference of group furnace operation and the upper limit of group furnace demand.In the smelting process,the power supply curve fluctuates violently,the product quality cannot be guaranteed,and the energy consumption per ton cannot be effectively reduced.This paper presents an on-line optimization method for energy efficiency scheduling of electric melting magnesium furnace based on demand prediction.This method takes the overall energy efficiency index of group furnace,i.e.the power consumption per ton,as the optimization objective,the upper limit of the maximum demand of group furnace as the constraint condition,considering the dynamic change of melting condition of the electric furnace,optimizes the power consumption of each electric furnace on-line,so as to improve the power utilization efficiency of the whole plant.The main research contents are as follows:First,combining the principle and characteristics of the electric melting magnesium furnace,this paper analyzes the energy efficiency scheduling optimization of the electric melting magnesium furnace from the perspective of dynamic environment and demand constraints.According to whether the demand exceeds the maximum limit,the energy efficiency optimization problem can be divided into single furnace optimization control and group furnaceoptimization scheduling,which can reduce the time cost and save electricity.Based on the dynamic change of scheduling decision variables and constraint conditions,the energy efficiency scheduling optimization model under dynamic environment was established to optimize the power utilization of group furnaces.Secondly,the premise of realizing the optimization of energy efficiency scheduling is to accurately predict the demand of group furnaces.The frequent switching of working conditions and the up and down movement of the electrode in the melting process lead to strong dynamic fluctuation of demand and strong nonlinear.In view of these characteristics,this paper proposes a method to predict demand based on Long Short-Tme Memory network combined with Attention mechanism,and verifies the superiority of this model by comparing different network structures.The demand prediction models of BP,LSTM,multi-layer LSTM,Gru and LSTM +attention network structures are designed respectively.The optimal parameters of LSTM +attention model are obtained by optimizing the super parameters such as activation function,hidden layer neuron,time step and small batch,and are applied to other network structures for comparative experiments.Finally,based on the industrial power consumption data of electric melting magnesium furnace collected on site,it is verified that the prediction performance of LSTM + attention model is the best,the error MAE is only 0.0564,1.4% higher than LSTM model;RMSE is only 0.076,2.5% higher than LSTM model;MAPE is only 0.02975,0.7% higher than LSTM model.Thirdly,the establishment of scheduling optimization model is the key step to realize the optimization of energy efficiency.The mathematical expression of power consumption per ton was established by studying the melting mechanism of electrofused magnesium furnace and analyzing the data.The power consumption per ton can comprehensively consider the production indexes such as total output and total power consumption,so as to maximize the output and minimize the energy consumption.Based on the mechanism and data analysis,the energy efficiency scheduling optimization model is established,which takes the single ton power consumption as the optimization goal,the demand limit of the group furnace as the constraint condition,and the power load in the group as the decision variable.The feasibility of the model is verified by the electric data of the electric melting furnace.Fourth,based on case-based reasoning(CBR)and particle swarm optimization(PSO)algorithm,combined with the melting characteristics of electrofused magnesium furnace,an energy efficiency scheduling optimization method based on demand prediction is proposed.According to the priority of working conditions,the decision-making rule base of electric melting furnace scheduling is established,and the initial case base is established based on the rule base and historical optimization solution.Based on case-based reasoning and historicalprior knowledge,the initial population of particle swarm optimization is generated,and the search process and results of particle swarm optimization algorithm are studied through case study,so as to improve the case library and improve the accuracy of case-based reasoning in the next shot.The guidance function of case-based reasoning(CBR)to particle swarm optimization(PSO)is further promoted,and the online cyclic interaction between CBR and PSO is realized.The proposed energy efficiency optimization method was verified by using the data collected from the industrial field of electrofused magnesium furnace,and compared with the effect of traditional manual decision making,so that the power consumption of group furnaces is saved by about 1.07%,the power consumption per ton is reduced by 6.1%,the power supply curve of group furnaces is optimized,and the energy utilization efficiency is improved.