基于混合差分进化算法的软测量时延参数估计  

Time delay estimation of soft-sensor model based on hybrid differential evolution algorithm

作　　者：王钧炎[1] 黄德先[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《化工学报》2008年第8期2058-2064,共7页CIESC Journal

基　　金：国家自然科学基金项目(60574072);国家高技术研究发展计划项目(2007AA04Z193);国家重点基础研究发展计划项目(2002CB3122002)~~

摘　　要：时延参数估计是系统控制与信号处理的关键问题。通过构造一个适当的适应度函数,将软测量系统的时延参数估计问题转化为一个多维非线性优化问题,然后利用混合差分进化算法的全局搜索能力求解该优化问题。对两个典型问题进行了仿真实验,仿真结果表明了混合算法的有效性和鲁棒性。以石油炼制工业中典型装置常压塔为例,对其一线航空煤油的闪点软测量进行了应用验证,结果表明,时延参数估计的引入大大提高了软测量模型的精度,证实了混合差分进化算法的有效性。Time delay estimation is one of the most important issues in system control and signal processing.Through establishing an appropriate fitness function,the time delay estimation of soft-sensor model could be formulated as a multi-dimension nonlinear functional optimization problem,which could be solved by hybrid differential evolution（HDE）algorithm.Numerical simulation results on two benchmarks demonstrated the effectiveness and robustness of HDE.Taking a crude atmospheric distillation unit in oil refinery as an example,a numerical application of the soft-sensor model for the kerosene flash point was made.The application results demonstrated that the accuracy of soft-sensor model was greatly improved with the introduction of time delay estimation.Besides,the effectiveness of the proposed HDE was confirmed.

关 键 词：混合差分进化 软测量模型 时延估计 偏最小二乘 

分 类 号：TQ021.8[化学工程] TP183[自动化与计算机技术—控制理论与控制工程]