基于继承思想的时变性电力系统暂态稳定预测  

Transient Stability Prediction of Time-varying Power Systems Based on Inheritance

作　　者：汤奕[1] 崔晗[1] 党杰 TANG Yi;CUI Han;DANG Jie(Southeast University,Nanjing 210096,Jiangsu Province,China;Central China Branch of State Grid Corporation of China,Wuhan 430077,Hubei Province,China)

机构地区：[1]东南大学,江苏省南京市210096 [2]国家电网公司华中分部,湖北省武汉市430077

出　　处：《中国电机工程学报》2021年第15期5107-5118,共12页Proceedings of the CSEE

基　　金：国家自然科学基金项目(51877037);国家电网公司华中分部科技项目(适应大规模强稀疏性新能源接入的受端电网特性分析与运行控制技术研究)。

摘　　要：从数据驱动的角度研究暂稳预测问题存在模型变迁和样本匮乏的困扰。针对该问题提出了一种基于继承思想的暂态功角稳定预测方法,该方法适用于暂稳数据增量或暂稳镜像变化的电网过渡阶段,填补了现有预测方法的空缺区间。分别针对电力系统暂稳样本增量的数据时变性和暂稳特性变化的暂稳镜像时变性特点,提出可计及现有预测模型参数的深度继承方法,以及可考虑样本与特征集可延拓性的广度继承方法。基于增量学习的深度继承利用新增暂稳样本,计算暂稳预测模型参数的变化量,降低模型更新所需训练时间。基于迁移学习的广度继承建立了源系统与目标系统的特征集与样本集迁移通道,实现目标系统在小样本条件下的暂稳预测模型构建。算例结果表明,该文所提方法能够反映时变电力系统的暂态稳定特性的动态变化,在样本匮乏的过渡阶段,具有速度和精度优势。From data-driven perspective,the research of transient stability prediction faces the difficulty of variant model and insufficient samples.To solve this problem,this paper proposed a method of transient power angle stability prediction based on inheritance,which fills the gap among the existing methods.Deep-inheritance was proposed to handle the time-varying dataset,and wide-inheritance deals with time-varying mirror of transient stability.Deep-inheritance was achieved through incremental learning,using the newly added transient stability samples to calculate the parameters of transient stability prediction model parameters,and reducing the training time required for model updates.Wide inheritance was based on the combination of feature transfer and sample transfer,which established the feature set and sample set transfer channel between source system and target system.It enabled the construction of the transient stability prediction model of the target system under the few samples condition.The test results show that the method can reflect the dynamic changes of the transient stability characteristics of time-varying power systems,and has the advantages in speed and accuracy under the condition of insufficient samples.

关 键 词：暂态稳定预测 时变电力系统 继承思想 增量学习 迁移学习 

分 类 号：TM712[电气工程—电力系统及自动化]