基于鲁棒LS-SVM的ARMA时序模型研究

对时序数据建模与辨识技术进行了分析,提出了使用鲁棒LS-SVM算法建立ARMA时序预测模型。该模型是在LS-SVM的约束条件中加入鲁棒特性和时序模型参数,使之在求解的过程中对孤立点与噪声不敏感,并且能准确地辨识时序模型参数。考虑到高炉的热状态的输入输出数据集间存在着复杂非线性时序上的关系,通过用基于鲁棒LS-SVM的ARMA模型预报铁水中硅的含量,从而达到了预测高炉热状态的目的。说明了该模型在对非线性时间序列预测精度和稳定性上具有明显的优越性,为稳定钢铁质量和生产工艺创造了良好条件。

Research of ARMA Time Series Model Based on Robust LS-SVM

Time series modeling and identification techniques were analyzed and the ARMA time series model based on robust LS-SVM algorithm was proposed. In the model, the robust character and time series model parameters have been added into constrain condition of LS-SVM. In the process of computation, the model is not sensitive to the outliers and noises and accurately identifies parameters of time series model. Considering the complex nonlinear relationship of time series between the input and output data sets, the ARMA model based on robust LS-SVM was used to predict the content of silicon in molten iron to gain the heat state of blast furnace. Finally, the experiments with practical data prove that the presented model has obvious superiority in the precision and robust of nonlinear time series prediction. Thus it provides the fine condition to improve quality of steel and stabilize manufacturing craftwork.