Predictive Control For Eaf Electrode Regulating System With Dead-zone

EAF steelmaking is a complex industrial process, and the performance of electrode regulating system is an important factor for product benefit. It is requested that electrode regulating system can regulate the distance from electrode terminal to steel fast and accurately according to arc variety in steelmaking process. Thus the stability of arc length can be guaranteed and the goal of high efficiency and low consumption can be arrived. Owing to nonlinearity and parameters time-variation in electrode regulating system, the regulating effect of conventional electrode regulator is not satisfying. It results in high electricity cost and low efficiency. Therefore it is very important to research high performance electrode regulator for improving steel quality and reducing electricity cost.Owing to characteristic parameters such as inherent frequency, hydraulic damp ratio and exterior disturb, EAF electrode regulating system is a nonlinear time-varying system with dead-zone. It is not an easily accomplished task to realize stable control of arc length. Dead-zone parameters are time-varying and parameters in linear system are time-varying owing to pressure etc. It is a rule to build a dead-zone inverse model as controller output compensator to counteract dead-zone when dead-zone parameters are known. For dead-zone in electrode regulating system, it is difficult to be measured and the regulating velocity is requested to be very quick in electrode regulating system.Conventional neural network such as BP learning algorithm is slow, quicker learning algorithm should be applied to identify dead-zone parameters on line, then compensate it.The electrode regulating system is simplified to the dead-zone and the three-order linear system in series in this thesis, then identifies and compensates dead-zone. In order to compensate dead-zone, additional control signal is added to the controller. After dead-zone is compensated, the electrode regulating system is a linear system.The obvious disadvantage of traditional PID is that it can only take current action to EAF and its regulating lags actual state change usually. In the thesis, direct GPC is applied to EAF electrode regulating system which varies quickly. Predictive model can be modified quickly and calculate corresponding control input when model do not matches.Simulation is performed by matlab in the end and the result shows the output can track the set value when system exists exterior disturb and change of parameter. It shows good adaptation and good prospect.