面向预测控制的有限阶跃响应模型辨识  

Predictive control oriented identification of finite step response model

作　　者：丁宝苍[1] 陈桥[1] 谢亚军[1] 

机构地区：[1]西安交通大学电子与信息工程学院,西安710049

出　　处：《控制与决策》2016年第11期2030-2036,共7页Control and Decision

基　　金：国家863计划项目(2014AA041802);国家自然科学基金项目(61573269);陕西省自然科学基金项目(2016JM6049)

摘　　要：针对工业预测控制软件开发中所需有限阶跃响应模型辨识问题,给出一种辨识算法.该算法包含数据坏值的标识与插值、数据滤波、辨识案件分组、积分型输出的处理,适合实际软件开发与工程实现.在模型的计算环节,采用经典的最小二乘辨识方法和递增的惩罚因子.主要贡献是将稳定型输出和积分型输出统一处理,采用增量数据辨识和增量算法,与实际应用的动态矩阵控制技术完全吻合.仿真算例验证了所提出算法的有效性.An algorithm for identifying the finite step response（FSR） model is given, which is utilized in the predictive control software development. This algorithm includes the stamp and interpolation of bad data, data filtering, identification case grouping and handling of the integral output, which is applied to the software development and engineering implementations. In calculating the FSR model, the classical least squares method and increasing penalty factors are applied.The contributions include an unified procedure for handling both stable and integral outputs, and the incremental method for both the data and the identification algorithm, which are consistent with the actually applied dynamic matrix control（DMC）techniques. The simulation example verifies the effectiveness of the proposed algorithm.

关 键 词：预测控制 最小二乘法 有限脉冲响应 增量模型 惩罚因子 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]