厂级振荡严重威胁工业生产过程的安全平稳运行。现有振荡检测与诊断机制存在响应滞后、诊断困难等问题，不利于流程工业的绿色、高效和智能化发展。根据过程监控先导理论，生产数据的时频分析有助于揭示厂级振荡的本征模态、阻尼特性、局部波动、传播方向及演化趋势，促进系统性的振荡研究。围绕厂级振荡智能预测与自动抑制命题，本项目以发展快速、高维时频分析为基础，开展创新性研究，拟完成以下关键技术：①提出新型多元时频分析，结合集成学习，研究过程数据自适应预测方法，实现厂级振荡的早期检测；②基于行波延时、因果分析与逻辑回归，研究多层面的振荡时空传播溯源技术；③研究系统降阶传递函数辨识方法，配置阻尼控制单元，抑制源回路主导振荡。最终形成较为完整的厂级振荡早期检测、溯源与抑制框架，并在云南电网下属的火力发电机组中进行方法验证。项目研究成果将为实现智能优化制造提供理论与技术支持，助力我国流程工业的安全稳定、绿色高效运行。

Plant-wide oscillation is one of the most common abnormal phenomenon that seriously threatens the safety and stability of the process. The existing mechanisms for oscillation detection and diagnosis show common problems in terms of delayed response and compromised diagnostic performance, which cannot meet the requirement for ensuring the green, efficient and intelligent operation of process industries. The pilot theory of process monitoring highlights that the time-frequency analysis of the production data helps to reveal the intrinsic modes, damping characteristic, local fluctuation, propagation direction and evolution trend of the plant-wide oscillations, which may promote the systematic research of industrial oscillations. .Motivated by this newly challenge on the intelligent forecasting and self-healing control of plant-wide oscillation, this project conducts innovative researches based on modern multivariate time-frequency analysis, tending to complete the following key technologies: (i) We propose novel multivariate time-frequency tools to realize the adaptive prediction of the process data for early detection of plant-wide oscillation, by combining with the ensemble learning approach. (ii) The multilevel time-space propagation analysis for oscillation source localization is proposed by leveraging the wave-delay feature, causality analysis and logistic regression. (iii) Aiming at suppressing the dominant oscillation in the source loop, this project presents time-frequency based reduced-order transfer function identification method for configuring the damping control unit..Finally, a joint framework for early detection, localization and damping of the industrial plant-wide oscillation will be formed, where the effectiveness and capability of the work will be demonstrated in the thermal power generating unit that belongs to the Yunnan power grid. This project is expected to provide theoretical and technical support for the development of the intelligent optimal manufacturing, which is of great importance for ensuring the safe, stable, green and efficient operation of China's process industries.