Research On Intelligent Electrode Control Strategy For Electric Arc Furnace With Continuous Charging

Continuous feed electric arc furnace is the key equipment for scrap steel smelting.This equipment includes continuous feed control system,carbon and oxygen injection control system,electrode control system and many other control systems.Among these control systems,the electrode control system is the most critical one.Fast and accurate control of the electrode position plays an important role in improving the working performance of the electric arc furnace,reducing energy consumption and shortening the smelting time.However,there are strong disturbances in the arc furnace smelting process,and reasonable electrode control strategies can effectively handle the electrode control environment with strong disturbances.Therefore,electrode control strategy has become the focus of electric arc furnace control research.With the continuous development of artificial intelligence and industrial computer technology,intelligent control technologies represented by intelligent methods such as neural networks and reinforcement learning have developed rapidly in the field of industrial control,bringing new ideas to the research on electrode control strategies.Aiming at the problem of strong interference during the arc furnace smelting process,this thesis proposes a reinforcement-learning-based intelligent electrode control strategy along with its overall framework considering the advantages of reinforcement learning method that it can interact with the environment and respond quickly.After researching the electrode reinforcement learning algorithm,DDPG(Deep Deterministic Policy Gradient)was selected as the algorithm of electrode reinforcement learning control strategy based on its advantages such as continuous output values,easy and fast convergece.The states and actions in the electrode reinforcement learning framework are also designed.Besides,the reward-and-punishment function is designed in segment by analyzing the control environment of the electrodes,and this design achieves the effect that the reward-and-punishment signal contains multiple environmental information.The necessity of establishing an electrode control environment is analyzed,and the electrode control environment is modeled by mechanism modeling.In addition,the arc model,power supply system environment and hydraulic system environment are also modeled respectively.The correctness and effectiveness of the established electrode control environment are verified by simulations.In the end,the stability of the electrode reinforcement learning control strategy is studied and analyzed by training the established framework,and the control effect of the trained control strategy is simulated.The simulation results show that the reinforcementlearning-based control strategy can stably control the electrode.In addition,this control strategy also improves the anti-interference ability and response speed compared with the traditional method,and it effectively solves the problem of strong interference during the arc furnace smelting process,thus verifying the rationality and effectiveness of the proposed control strategy.