摘要
准确、快速的暂态稳定评估对电网安全运行至关重要。但现有方法未充分挖掘电网暂态数据的时空特性信息,限制了模型的评估性能。文中提出了一种基于时间卷积网络(TCN)和图注意力网络(GAT)的暂态稳定评估方法。该方法仅以量测母线电压幅值和相角数据作为输入,凭借GAT可以处理图数据并建立电网拓扑连接关系的优点和TCN特有的因果空洞卷积运算特性,自动从暂态数据中提取出空间特征和时间特征,进而实现对系统暂态稳定性的准确评估。此外,采用改进的焦点损失函数作为模型训练目标,可以动态适应训练过程中模型对难易样本的判别界限且自适应处理样本不均衡问题,减少了对失稳样本错分类的现象,同时还提高了全局准确率。IEEE 39和IEEE 145节点系统仿真结果表明,所提方法在响应时间上具有优越性,并且在拓扑变化和数据存在噪声情况下都具有较强的泛化性和鲁棒性,满足在线评估的准确性与快速性要求。
Accurate and fast transient stability assessment is very important for the safe operation of the power system.However,the existing methods do not fully exploit the spatial-temporal characteristics of the power grid transient data,which limits the assessment performance of the models.This paper proposes a transient stability assessment method based on a temporal convolution network(TCN)and graph attention network(GAT).This method only takes the measured bus voltage amplitude and phase-angle data as the input.With the advantage that GAT can process the diagram data and establish the power grid topology connection relationship and the unique causal null convolution operation characteristics of TCN,the spatial and temporal characteristics are automatically extracted to realize the accurate assessment of transient stability.In addition,the improved focus loss function is used as the model training target,which can dynamically adapt to the discriminant boundary of the difficult and easy samples in the training process and adaptively deal with the problem of sample imbalance,thus reducing the phenomenon of misclassification of unstable samples and improving the global accuracy.The simulation results of the IEEE 39-bus system and IEEE 145-bus system show that the proposed method is superior in response time,has strong generalization and robustness in the case of topological structure changes and data noise,and meets the accuracy and rapidity requirements of online transient stability assessment.
作者
张亮
安军
周毅博
ZHANG Liang;AN Jun;ZHOU Yibo(Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology,Ministry of Education,Northeast Electric Power University,Jilin 132012,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2023年第7期114-122,共9页
Automation of Electric Power Systems
基金
国家自然科学基金资助项目(51877034)。
关键词
暂态稳定
评估方法
深度学习
图注意力
时间卷积
焦点损失
时空特征
transient stability
assessment method
deep learning
graph attention
temporal convolution
focus loss
spatial-temporal characteristics
