| 引用本文: | 张异浩,韩松,荣娜.基于CNN-Informer和DeepLIFT的电力系统频率稳定评估方法[J].电力自动化设备,2025,45(7):165-171 |
| ZHANG Yihao,HAN Song,RONG Na.Frequency stability assessment method of power system based on CNN-Informer and DeepLIFT[J].Electric Power Automation Equipment,2025,45(7):165-171 |
|
| 摘要: |
| 为解决扰动发生后电力系统频率稳定评估精度低且预测时间长的问题，提出了一种电力系统频率稳定评估方法。该方法改进层次时间戳机制，有效捕捉了频率响应在不同时间尺度下的相关性；利用深度学习重要特征技术对输入特征进行筛选，简化了数据维度并提升了模型的训练效率和预测性能；结合卷积神经网络与Informer网络，基于编码器与解码器的协同训练，构建适用于多场景的频率稳定评估框架。以修改后的新英格兰10机39节点系统和WECC 29机179节点系统为算例，仿真结果表明，所提方法在时效性和准确性方面具有显著的优势，并在多种实验条件下展现出良好的鲁棒性和适应性。 |
| 关键词: 电力系统 频率稳定评估 深度学习 时序数据 层次时间戳 蒸馏机制 卷积神经网络 |
| DOI：10.16081/j.epae.202502009 |
| 分类号:TM712;TP18 |
| 基金项目:贵州省优秀青年科技人才项目([2021]5645)；贵州省科技支撑计划项目([2023]290,[2023]329)；贵州省科学技术基金资助项目([2021]277) |
|
| Frequency stability assessment method of power system based on CNN-Informer and DeepLIFT |
|
ZHANG Yihao, HAN Song, RONG Na
|
|
College of Electrical Engineering, Guizhou University, Guiyang 550025, China
|
| Abstract: |
| To address the issues of low accuracy and long prediction time in power system frequency stability assessment after disturbances, a frequency stability assessment method of power system is proposed. The method improves the hierarchical timestamp mechanism and effectively captures the correlation of frequency response in different time scales. Deep learning important feature technique is used to select input features, the data dimension is simplified and the training efficiency and prediction performance of the model are enhanced. By integrating convolutional neural network with Informer network, and employing encoder-decoder cooperative training, a frequency stability assessment framework suitable for multi-scenario applications is constructed. Experimental results on the modified New England 10-machine 39-bus system and the WECC 29-machine 179-bus system demonstrate that the proposed method has significant advantage in terms of time efficiency and accuracy, and exhibits robust performance and strong adaptability under various experimental conditions, providing valuable references for frequency stability assessment and control in actual power system operation. |
| Key words: electric power systems frequency stability assessment deep learning chronological data hierarchical timestamp distillation mechanism convolutional neural network |
