论文部分内容阅读
高炉炼铁是一种具有强时延性的非线性系统,本文针对高炉炼铁过程中含硅量[Si]的动态预测问题,进行了数据挖掘与模型建立。首先对数据进行了预处理,包括3σ原则验证与归一化处理。同时[Si]预测问题可看做一个多输入单输出的多元时间序列问题,本文选取了优化PSO-BP神经网络算法来得到多元时间序列模型。在一步预测模型的基础上,通过将一步预测结果作为二步预测的变量的方法,进一步建立了[Si]的二步预测模型。最后结合模型的预测图以及各类性能指标对[Si]预测模型进行了评估分析。
Blast furnace ironmaking is a kind of nonlinear system with strong time delay. In this paper, data mining and model establishment are carried out for the dynamic prediction of silicon content [Si] in blast furnace ironmaking process. First of all, the data were preprocessed, including 3σ principle verification and normalization. At the same time [Si] predicts the problem can be seen as a multi-input single output multiple time series problem, this paper selects the optimized PSO-BP neural network algorithm to get multiple time series model. Based on the one-step prediction model, the two-step prediction model of [Si] is further established by taking the one-step prediction as the variable of two-step prediction. Finally, the forecasting model of [Si] is evaluated with the prediction model of the model and various performance indexes.