摘要
为进一步提高电力系统暂态电压稳定评估模型的特征提取能力和模型在系统拓扑结构发生变化时的适应性,提出一种将改进的卷积神经网络与迁移学习相结合的方法。首先,在卷积神经网络的卷积层后插入卷积块注意力模块,对输入的数据从通道和空间两个独立的维度依次提取特征,提高卷积神经网络对系统暂态电压状态的识别能力。然后,将该模块与微调技术相结合,提高模型在系统拓扑结构改变时的在线更新速度。最后,算例分析验证了所提模型的有效性。
To further improve the feature extraction capability of a short-term voltage stability assessment model for power system and its adaptability when the system topology changes,a method combining an improved convolutional neural network(CNN)with transfer learning is proposed.First,a convolutional block attention module(CBAM)is inserted after the convolution layer of CNN to extract features from the input data in two independent dimensions of channel and space sequentially,thus improving the capability of CNN to recognize the system’s short-term voltage state.Then,the module is combined with the fine-tuning technology to improve the model’s online update speed when the system topology changes.Finally,the analysis of numerical examples verifies the effectiveness the proposed model.
作者
李欣
柳圣池
李新宇
陈德秋
鲁玲
郭攀锋
LI Xin;LIU Shengchi;LI Xinyu;CHEN Deqiu;LU Ling;GUO Panfeng(College of Electrical Engineering&New Energy,China Three Gorges University,Yichang 443000,China;Hubei Provincial Engineering Research Center of Intelligent Energy Technology,China Three Gorges University,Yichang 443000,China;China Three Gorges Corporation,Yichang 443631,China;Xianning Power Supply Company,State Grid Hubei Electric Power Co.,Ltd,Xianning 437100,China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2024年第4期59-67,75,共10页
Proceedings of the CSU-EPSA
关键词
深度学习
卷积神经网络
暂态电压稳定评估
卷积块注意力模块
迁移学习
deep learning
convolutional neural network(CNN)
short-term voltage stability assessment
convolutional block attention module(CBAM)
transfer learning
