| Electric energy is the main sources of electric ARC furnace steel-making, appointment account for 60 ~ 70%. The optimization technology lies in the ability bringing the electric arc stove transformer into play sufficiently, the electricity operation using electric arc steel-making process reaches an optimization, thereby, reaches rise smelt efficiency , shortens a duration of heat , saves general electric energy target. The key technology is to build the development electricity operation model , development model is getting more reliable as vivid , analyzing a dynamic data more , the upper optimization of basis studies optimum state in approaching reality right away increasingly here in the research realizing an optimization.The tradition control system is basis static state the electricity characteristic property curve working out controls electric arc steel-making, ensures that the system tracking curve works. During the period of steel-making practice works, because of the ARC existence, that is nonlinearities, produce second high harmonics in main circuit, there is quite difference between the operation curve and actual operational state of system that feasible static state calculation. Unable accurate relation between reflecting each electric of system parameter, control system cannot reach the very good request. For this purpose bring out optimization control system, is on building the development electricity model basis, analysis reaches the optimum working spot of development by the fact that the optimization various studies achievement.Our research use BP NN to modeling the connection of dynamic electric, making use of neural networks to be able to study , distinguishing and adapting to smelting process middle furnace loading characteristic property ceaselessness, that the working adjusting a furnace's enactment , studies by the optimization development, makes a furnace work all the time within the most rational range , can realize more fine and capability than the tradition technology, gain an advanced technology index more. Neural networks great majority adopt the BP network model, have the structure simplicity, and good adaptability, strong broad-relation ability, stable state of work, can use accuracy approaches arbitrarily the function merit, has gained broad application arbitrarily. The article will also build a model mainly to BP neural networks and Least square method discusses. |
