基于凸优化的闭环辨识测试信号设计  

Convex Optimization Based Testing Signal Design for Closed-loop Identification

作　　者：段晓宁[1,2] 林文益[1,2] 谢磊[1,2] 

机构地区：[1]浙江农业商贸职业学院财会金融系,绍兴312088 [2]浙江大学工业控制技术国家重点实验室智能系统与控制研究所,杭州310027

出　　处：《控制工程》2017年第7期1297-1302,共6页Control Engineering of China

基　　金：国家自然科学基金资助项目(60421002)

摘　　要：辨识过程中测试信号的设计与选取是影响最后辨识结果准确性的重要因素,针对PID闭环回路进行了最优测试信号设计研究,提出了以测试信号功率为约束,最小化参数协方差矩阵的闭环最优测试信号设计方法,通过在设定点上添加测试信号,保证系统安全与质量稳定。在频域范围内构造凸优化命题进行求解并将测试信号转化为余弦信号叠加表达式。最后使用Monte Carlo仿真测试证明该测试信号在同等能量的前提下能比白噪声测试信号和随机二进制测试信号得到更好的模型。Testing signal design is of critical importance for the model quahty m system ldenntlcanon, lneoptimal test signal design for a closed-loop system with PID controller is studied. A new closed-loop testingsignal design approach is developed based on the minimization of parameter covariance matrix which issubject to the constraint of testing signal energy. We add the test signal at the set point to ensure the systemsafety and quality stability. Then we construct and solve the corresponding SDP problem in frequencydomain and use the cosine superimposed to characterize the result. At last, Monte Carlo simulations showthat at the same test energy, the optimal signal can get better model than the white noise signal and randombinary signal.

关 键 词：闭环辨识 最优信号设计 凸优化 

分 类 号：TP802[自动化与计算机技术—检测技术与自动化装置]