时延系统中T-LSSVR动态软测量建模方法研究

Dynamic Soft Measurement Modeling Method of T-LSSVR in Systems with Time Delay

针对工业过程中由于系统存在延时导致软测量模型难以建立、模型精度偏低等问题,提出将系统延时(T)与最小二乘支持向量回归机(LSSVR)相结合,构建一种基于T-LSSVR的动态软测量建模方法;该方法在建模过程中利用互相关函数与一阶广义差分算法辨识得到"静态响应延时"和"动态响应延时",通过软测量手段对变量进行预测以实现辅助变量对主导变量的最佳估计。对某化工企业具有此类双延时性质的系统进行实验,实验结果表明该建模方法在动态和稳态数据预测方面都有良好的预测效果。

For the delay problem in industrial process system,which lead to not soft sensor modeling in real-time or the lower accuracy of measurement,a modeling method for dynamic soft measurement based on a new algorithm( TLSSVR) which combines the system delay( T) and least squares support vector regression( LSSVR) was presented. In order to achieve the best estimate of the auxiliary variables,the'static response delay'and'dynamic response delay'can be identified by cross-correlation function and the first order difference algorithm during modeling to the leading variables by the means of soft measurement to predict variables in the method. The soft measurement model was applied to a system with such double-delayed nature of a chemical company and the results show that its prediction achieved good results in terms of both dynamic and steady-state data.