This paper presents a new method of instability detection and subsequent stabilization of the power system network using a proposed multi-shot remedial action scheme (RAS) that can extemporaneously detect critical generators based on the dynamic states of generator computed from the terminal phasor measurement units. The instability detector is a moving window classifier that predicts impending instability using rate of change of individual generator transient energy indices evaluated from the d–q axis voltage as well as conventional severity indices based on generator angle and frequency. A comparative performance analysis of a spectral feature based ensemble decision tree classifier with a multivariate long short-term memory network is also presented. The proposed RAS identifies critical generators through individual machine transient energy formulation and recursive coherency matrix, evaluated solely from system-wide generator dynamic states, and maintains stability by tripping adaptively the run-away generators. Performance evaluation of the proposed scheme has been made on IEEE 39-bus network and it has been demonstrated that the proposed RAS is robust with regards to instability prediction and it can effectively identify critical generators and stabilize the network by tripping the same.

中文翻译：

---

基于 PMU 信号响应的 RAS 通过运行中识别和失控发电机脱落来缓解不稳定性

本文提出了一种新的电力系统网络不稳定检测和后续稳定方法，该方法使用提出的多发补救措施方案 (RAS)，该方案可以根据终端相量测量单元计算的发电机动态即时检测关键发电机。不稳定检测器是一个移动窗口分类器，它使用从 d-q 轴电压评估的单个发电机瞬态能量指数的变化率以及基于发电机角度和频率的常规严重性指数来预测即将发生的不稳定。还介绍了基于频谱特征的集成决策树分类器与多元长短期记忆网络的比较性能分析。提议的 RAS 通过单个机器瞬态能量公式和递归相干矩阵识别关键发电机，仅从系统范围的发电机动态状态进行评估，并通过自适应跳闸失控发电机来保持稳定性。已在 IEEE 39 总线网络上对所提出的方案进行了性能评估，并且已经证明所提出的 RAS 在不稳定预测方面具有鲁棒性，并且可以有效地识别关键发电机并通过跳闸来稳定网络。