带时延估计的时间差分PLS软测量建模方法

针对实际工业过程中普遍存在的非线性和时延,提出一种带时延估计的时间差分偏最小二乘PLS软测量建模方法。该方法将过程变量的差分值作为训练样本建立PLS模型,在充分保留PLS的优点的同时,降低过程非线性对建模精度的影响。另外,基于差分进化算法估计时延,得到了时延估计的全局最优解,有助于进一步提高模型精度。针对常压塔常二线干点的建模结果验证了该方法的可行性和有效性。

Soft Measurement Modeling Method Based on Time Difference PLS with Time Delay Estimation

Aiming at the common issues of nonlinearity and time delay in actual industrial processes, a soft measurement method based on time difference partial least squares （PLS） with time delay estimation is presented. This method employs time difference values of process variables as training sample to construct PLS model, which can reduce the effects of process nonlinearily on the modelling accuracy with fully retaining the advantages of PLS simultaneously. In addition, global optimal solution is obtained based on the estimated time delay with differential evolution algorithm, which can help to further improve the model accuracy. The feasibility and effectiveness of the proposed method is verified with modeling result for normal pressure column.