基于NARX神经网络的高炉冶炼过程Si元素的预测与智能控制

Prediction and Intelligent Control of Si Element in Blast Furnace Smelting Process Based on NARX Dynamic Neural Network

使用NARX动态神经网络建立高炉冶炼过程中Si元素的单步预测与双步预测模型,分析使用NARX神经网络建模在过程工业中进行动态控制的可行性。通过实验结果发现,单步预测的命中率(误差±0:1)达到了近100%,而方向预测的正确率达到了80%。同时,双步预测的命中率也达到了近100%,而对变化方向的预测的准确率达到了90%,可以看出运用NARX神经网络对高炉冶炼过程Si元素进行预测控制的可行性并具有较高的预测精度。

NARX dynamic neural network was used to establish the step-ahead prediction network and two-step prediction network of Si element in blast furnace smelting process, and the feasibility of dynamic control in process industry using NARX neural network modeling was analyzed. The experimental results showed that the hit rate（error ±0.1） of the single step was almost 100%, and the correct rate of the direction was 80%. At the same time, the hit rate of the two-step prediction was nearly 100%, and the accuracy of the forecasting direction was 90%. Therefore, the NARX dynamic neural network can be applied in the dynamic process control to get a good performance of prediction.