一种自适应动态软传感器智能化建模方法

An Intelligent Modeling Method of Adaptive and Dynamic Soft Sensor

软传感器技术是解决流程工业工程中关键参数实时测量困难问题的有效手段。然而,流程工业过程呈现出十分复杂的特性,包括长流程以及大时间常数引起的时延与动态特性、内外部变化因素引起的时变特性等,导致建立高精度的软传感器模型十分困难。为此,论文提出一种基于人工智能技术的智能化自适应动态软传感器建模方法:以滑动时间窗框架为基础,解决过程的非线性与时变特性;以动态偏最小二乘法解决动态特性;同时,差分进化算法实现对所有参数的智能估计,从而减少人工干预并实现最优性。论文提出的方法采用公开的工业数据集进行了应用验证。

Soft sensing technology has proven to be an effective solution to the difficulties in measuring industrial key parame⁃ters in real-time.However,process industries exhibit complex characteristics,including the time delays and dynamic characteris⁃tics caused by long transportation distance and large time constant,as well as the time-varying characteristics resulting from change⁃ful internal and external factors,which makes the development of a soft sensor with high predictive accuracy difficult.Therefore,this paper proposes an artificial intelligent technology-based intelligent method for developing adaptive and dynamic soft sensor models.In the proposed method,the moving window framework is employed to deal with nonlinear and time-varying issues,the dy⁃namic partial least squares model is used to handle process dynamics.In addition,the differential evolution algorithm is adopted to realize the optimization of the whole algorithmic parameters intelligently,such that manual work can be saved,and most importantly the optimality can be guaranteed.A performance of the proposed method is evaluated by real-life industrial dataset for application verification.