An Intelligent Control Of Electric Arc Furnace Based Neural Network

Electric arc furnace (EFA) system is a complex plant, which exits serious non-linearity, time changing, strong couple. The constant impedance control method, which base on classical control theory has been used widely. In this paper the shortcomings of the constant impedance control method has been analyzed , in accordance with EFA system' s characteristics, requests of design and developing direction of EFA control system an intelligent control method has been proposed. Based neural network the model of the EFA is established, which can pre-estimate the echo of control variable of EFA. By analyzing the result of pre-estimate the control variable can then be adjusted in real time.The main works of the paper as follows:1. Based neural network the model of EFA has been established. Since EFA is a complex plant, it is difficult to establish the model. This problem has been solved through two-step. 1)Sample data has been used to train a BP neural network based genetic algorithm (GA), the model is a static model.2) The reformed BP algorithm bas been used to adjust the model in real time.2. Some research works on the multi-step pre-estimated neural network model has been done.3. In accordance with EFA system' s characteristics, three kind control schemes of EFA are designed. Simulation results show the effectiveness and robustness of the schemes.4. Combine the constant impedance controller with the schemes mentioned above. An new intelligent controller of EFA is proposed. The results prove that the scheme increases the stability of electrode apparently, current fluctuating has been reduced, speed of system echo has been improved. The system has fine self-adaptive ability and fine ability to resist interference.