基于改进收敛交叉映射的厂级振荡源诊断

Location of Plant⁃Level Oscillation Source Based on Improved Convergent Cross Mapping

周学均 陈小强 谢磊

Xuejun Zhou Xiaoqiang Chen Lei Xie

中电国际胡布发电有限公司, 巴基斯坦 卡拉奇 74200 中电华创电力技术研究有限公司，江苏 苏州 215123 浙江大学 智能系统与控制研究所，浙江 杭州 310027

China Power Hub Generation Company，Karachi 74200，Pakistan CLP Huachuang Power Technology Research Co. ，Ltd. ，Suzhou Jiangsu 215123，China Institute of Cyber?Systems and Control，Zhejiang University，Hangzhou Zhejiang 310027，China

由于噪声、周期性、非线性和非平稳性的干扰，现有的大多数因果分析方法在工业过程控制系统的应用中往往是不可靠和不准确的。为提高厂级振荡源定位的性能，提出了一种基于改进收敛交叉映射的因果关系检测方法。首先，指出噪声和周期性对因果关系检测的不利影响。其次，将经验模态分解和去趋势波动分析相结合，实现了振荡信号去噪。然后，通过奇异谱分析有效地去除信号的周期性。利用收敛交叉映射可以对去噪去周期后的信号进行分析，进而准确地定位厂级振荡的源头。仿真结果表明，所提方法能提高过程控制系统中厂级振荡源定位的准确性。

Due to the interference of noise, periodicity, nonlinearity and nonstationarity, most of the existing causal analysis methods are often unreliable and inaccurate in industrial process control systems. In order to improve the plant-level oscillation source location performance, a causal relationship detector based on improved convergence cross mapping was proposed. First, the adverse effects of noise and periodicity on causality detection were pointed out. Then, the empirical mode decomposition (EMD) and the trend fluctuation analysis were combined to realize the oscillation signal denoising. The periodicity of the signal was effectively removed by singular spectrum analysis. The denoising and periodicity?removing signal was analyzed by using convergent cross mapping, and then the source of plant level oscillation could be accurately located. Simulation results show that, the proposed method can improve the accuracy of plant?level oscillation source location in process control system.

国家自然科学基金项目(62073286)

10.3969/j.issn.1672-6952.2021.05.015