Research On Energy-efficient Scheduling Method Of Data-driven Electro-fused Magnesium Furnace

Electric furnaces are widely used in steelmaking,smelting glass,smelting non-ferrous metals,etc.,and play major roles in national defense and economic development.At the same time,electric furnace is also a high-energy consumption and material consumption equipment.By optimizing its power supply curve,it can improve electrical efficiency and reduce the cost of electricity,which has important research significance and application value for improving enterprise economic efficiency and achieving energy saving.At present,the single-furnace power consumption optimization strategy is adopted by enterprises,and the maximum demand constraint of group furnaces can't be considered.As a result,the electro-fused magnesium furnace is frequently opened and cut off,which reduces the electrical efficiency.In order to solve this problem,supported by National Natural Science Foundation of China this thesis is part of the project "Research on Dynamic Multi-objective Closed-loop Optimization Method for Flexible Load Scheduling in Electro-metallurgical Process"(61603088),proposes an energy-efficient scheduling method for the electric magnesium melting furnace,which takes the electrical efficiency of the group furnace as the goal,and schedules the electric load of the electric magnesium melting furnace from two aspects of space and time under the premise of satisfying the constraint of the maximum demand.The main research work includes the following aspects:1)This thesis analyzes the energy-efficient scheduling problem of electro-fused magnesium furnace from two angles of time and space: First,the electricity demand of electric furnace fluctuates sharply during the smelting process,resulting in unstable production,reduced electrical efficiency,and electric furnace in time.Distributing with electric load can cut peaks and fill valleys and reduce the cost of electricity required by enterprises.Second,the average load of group furnaces is constrained by the maximum demand,and the energy is optimized according to the difference in electricity efficiency between different electric furnaces to realize the group.The overall electrical efficiency of the group furnaces is optimized.2)The power demand for electric furnace changes dynamically due to changes in smelting conditions,so accurate prediction of the smelting conditions is the basis for energy-efficient scheduling.The smelting conditions are cyclical and are affected by uncertain factors such as human operation and environmental changes.The traditional rule-based reasoning method can't effectively use uncertain information and can't predict future changes in conditions.In order to solve this problem,a method based on confidence rule reasoning is proposed to predict the smelting conditions of electro-fused magnesium furnace.By analyzing smelting condition data and relevant expert experience,characteristic vectors such as current setting error and absolute value of current change were extracted.Characteristic vectors and prediction results of smelting conditions were taken as the precondition and result attributes of the model respectively to establish the prediction model of smelting condition.The actual production data of the electric fused magnesium enterprise were used to confirm the proposed method,and the results showed that the prediction accuracy of the smelting condition was more than 95%.3)The prediction of group furnaces demand is the key to realize the energy-efficient scheduling.The variation of demand is influenced by the elevation and fluctuation of electrodes and the dynamic working conditions,and has strong non-linear,dynamic and random fluctuation characteristics.According to the characteristics of the demand change,a prediction method of electro-fused magnesium furnace demand based on recurrent neural network is proposed,The forecasting models of different network structures such as LSTM,GRU,RNN,bi-LSTM,dual-LSTM and triple-LSTM are designed,and discusses the different hyper-parameters,activation functions and network structure on the properties of demand forecasting model under the influence,the proposed method is compared with the least squares support vector machine(LSSVM)demand forecasting model.The authentic production data of the electric fused magnesium enterprise were used to verify the demand forecasting method,and the autocorrelation analysis of the error sequence was carried out.The experimental results show that the forecasting accuracy of electro-fused magnesium furnace demand based on LSTM is the highest,MAE is 5.72%,RMSE is 7.8% and MAPE is 2.996%.4)Based on the above research,combined with heuristic rules and rolling decision ideas,the energy-efficient scheduling strategy of electro-fused magnesium furnace based on heuristic rules is proposed.By predicting the dynamic characteristics of the smelting conditions,determining the priority of the electric furnace scheduling,and combining the scheduling priority with the expert experience to establish an energy-efficient scheduling rule base.The real production data of electric fused magnesium enterprise were used to verify the energy-efficient scheduling method.The results show that the energy-efficient scheduling method of electro-fused magnesium furnace can significantly reduce the number of trips and power failures,and the load rate increases by 4.94%,ensuring the stability of production and improving the electrical efficiency.