计及同步电机饱和影响的小干扰稳定评估方法
Small disturbance stability evaluation method considering the saturation impact of synchronous machine
刘川 1李登峰 2刘育明 2徐瑞林 2李小菊 2黄淼1
作者信息
- 1. 重庆邮电大学 自动化学院,重庆 400065
- 2. 国网重庆市电力公司电力科学研究院,重庆 401123
- 折叠
摘要
目前,基于机器学习的电力系统小干扰稳定评估方法往往只利用稳态潮流数据来构建机器学习模型,未充分考虑同步电机饱和特性这一影响小干扰稳定分析结果的重要因素.一旦同步电机由于改造、更换等因素导致其饱和特性发生变化,反映小干扰稳定性的物理量(如特征值、最小阻尼比等)可能也会发生变化,但基于潮流数据得到的小干扰稳定结果却不会改变,无法准确反映小干扰稳定性.为此,提出一种计及同步电机饱和影响的小干扰稳定评估方法.将卷积神经网络(convolutional neural network,CNN)和图卷积神经网络(graph convolutional neural net-work,GCN)融合,构建CNN-GCN模型来提升机器学习模型的特征提取能力.该模型同时考虑同步电机饱和系数和稳态潮流数据,实现对最小阻尼比的预测.针对IEEE14节点系统的算例结果验证了所提方法的有效性和优越性.
Abstract
Currently,the small disturbance stability evaluation methods based on machine learning only use steady power flow data to build machine learning models.Therefore,the saturation characteristics of synchronous machine,an important factor affecting the analysis results of small disturbance stability,are not fully considered.Once the saturation characteristics of a synchronous machine change due to retrofit,replacement and other factors,the physical quantities reflecting the stability of small disturbance,such as characteristic values,minimum damping ratio,may also change whereas the corresponding analysis results based on power flow data remain unchanged.Thus,the stability of small disturbance is not accurately reflected.To address the issue,this paper proposes a small disturbance stability evaluation method with full considerations of the saturation impact of synchronous machines.The method integrates Convolutional Neural Network (CNN)and Graph Convolutional Neural Network (GCN)and then a model called CNN-GCN is built to improve the feature extraction ability of machine learning model.In this model,saturation coefficient of machine and steady-state power flow data are employed to predict the minimum damping ratio.The effectiveness and superiority of the proposed method are validated by the IEEE14-node system.
关键词
同步电机饱和/小干扰稳定/机器学习/融合模型Key words
synchronous machine saturation/small signal stability/machine learning/fusion model引用本文复制引用
基金项目
国网重庆市电力公司科技项目(2023渝电科技11)
国网重庆市电力公司科技项目(SGCQDK00DYJS2310173)
出版年
2024
国家科技期刊平台
NSTL
万方数据