论文部分内容阅读
针对氧化铝蒸发过程的工业现场出口料液浓度在线检测困难、操作参数具有时变性以及传统离线预测所存在的不足等特点,提出了一种多输入多输出系统的自适应加权最小二乘支持向量回归,并用于氧化铝蒸发过程出口料液浓度的在线预测。该方法根据模型预测效果自适应在线调整建模的训练样本集,利用主元分析提取主元作为分段加权支持向量回归模型的输入,采用网格搜索和交叉验证法对多输入多输出模型参数进行优化。采用工业现场的实测数据进行实验分析,计算结果表明:该方法能够很好地在线预测氧化铝蒸发过程出口料液浓度,相比基于最小二乘支持向量回归以及基于BP神经网络的浓度预测模型,该方法具有更高的预测精度和更好的泛化性能,满足实际工业生产在线优化控制要求。
In view of the difficulty of on-line detection of liquid concentration of industrial outlet in alumina evaporation process, the variability of operating parameters and the shortcomings of traditional offline prediction, an adaptive weighted least square support vector for multi-input multi-output system Regression, and is used for the on-line prediction of the outlet feed concentration in the alumina evaporation process. The method adjusts the training sample set adaptively on-line according to the prediction effect of the model, uses the principal component analysis (PCA) to extract the principal component as the input of the subsection weighted support vector regression model, and uses grid search and cross-validation method to test the MMI parameters optimize. The experimental data of industrial field were used for the experimental analysis. The calculated results show that the proposed method can predict the outlet liquid concentration of alumina evaporation process well. Compared with the regression model based on least squares support vector regression and the BP neural network based concentration prediction model, The method has higher prediction accuracy and better generalization performance, and meets the actual online optimization control requirements of industrial production.