Abstract: In accordance with such characters of the electrode control system in ladle furnace as the high non-linearity, time-variant, uncertainty of the model, output response time delay serious, and multivariable input and output coupling, an internal model control strategy based on real-time identification on line by neural network was presented. The control strategy applies neural network decoupline control and the chaos algorithm to the improved BP algorithm and speeds up the training of neural network. The validity of the control strategy is verified by simulation analysis.