摘要
暂态稳定实时性判别是响应式稳定控制的核心,新能源和直流输电并网下电网的结构、运行和响应特征复杂程度骤增,现有实时性判别方法的准确性和泛化性面临挑战。该文针对含新能源和直流输电的复杂电网暂态稳定性,提出了一种基于可量测暂态能量特征的自适应逻辑推理判稳方法,实现无电网模型依赖且具有可解释性的实时性稳定性判别。首先,基于发电机转子运动的能量关系,结合特勒根定理构建了面向复杂电网响应信息的能量函数并论证了其守恒性;然后,根据系统动能-势能能量转换特征定义了稳定预判因数,并结合最大功角差构成了判稳关键特征量,提出了基于自适应模糊推理神经网络(ANFIS)的关键特征量与稳定状态之间的映射模型,实现了暂态稳定性实时推理评估;最后,在简单系统中量化分析了关键特征量与系统稳定性间的关系,并在修改后含新能源与直流输电的IEEE10机39节点系统中验证了该文所提方法的有效性与泛化性。
Real-time transient stability assessment is the core of response-driven stability control.With renewable energy and DC transmissions connecting to power systems,the complexity of grid’s structure,operation and characteristics increases significantly,which have challenged the effectiveness and accuracy of existing stability assessment methods.For real-time stability assessment in the complex power system including renewable energy and DC transmissions,an adaptive network based fuzzy inference system(ANFIS)based model-free and interpretable stability assessment method was proposed by using measurable transient energy features.First,based on the energy relation of generator motion equation and the Tellegen’s theorem,the transient energy function using the response information of the complex grid was constructed and its energy conservation was also validated.Then,according to the conversion characteristic between kinetic energy and potential energy,the stability prediction factor was defined,which was combined with maximum angle deviation to obtain the key features of the stability judgment.A mapping model between the key features and stability status based on ANFIS adaptive logic reasoning was proposed to realize real-time prediction of transient stability.Finally,the quantitative analysis of the relationship between the key features and system stability status was illustrated in a simple system,and the effectiveness and generalization of the proposed method were verified in the modified IEEE 10-machine 39-bus system with new energies and DC transmissions.Firstly,based on the energy relation of the generator rotor movement and the Tellegen’s theorem,a refined energy function model was developed by using the response information of a complex grid,and the energy conservation property of the model was proven.The energy function model can depict the energy response characteristics of various components in the system.By introducing the virtual impedance branch,the potential energy of each component can be conveniently calculated using the pre-fault equilibrium point of power system,which does not depend on the component model and its parameters,showing good generality.Secondly,the conversion characteristics between kinetic energy and potential energy were analyzed under stable and unstable conditions,and then a stability prediction factor(SPF)were defined at the maximum potential energy point.Together with the maximum angle deviation(MAD)at the same point,the critical features SPF and MAD for stability assessment were obtained,which have strong representational ability to indicate the stable/unstable status of the system.Finally,a response-driven stability discrimination model was constructed with the key features SPF and MAD.The stability discrimination model was established based on logical reasoning ANFIS,which can obtain an interpretable mapping model between the key features and the stable/unstable status of the system.This model enables real-time transient stability assessment with interpretability.This paper quantitatively analyzed the relationship between the key features(SPF and MAD)and system stability status in a modified simple IEEE 9-bus system,and evaluated the effectiveness and generalization of the proposed stability discrimination model in a modified IEEE 10-machine 39-bus system with new energy sources and DC transmissions.The simulation results demonstrated that the proposed key features SPF and MAD,showing strong physical attributes associated with the system stability,can effectively reflect the system stability status.Compared to other AI stability discrimination models,the proposed data-driven ANFIS stability discrimination model had a higher accuracy of stability discrimination in a complex power grid.Thus,it can be effectively applied in subsequent emergency response-driven stability control.The conclusions of this paper are given as follows:(1)The key features SPF and MAD,extracted from the constructed energy function model,have a strong correlation with the system stable/unstable status.They can directly characterize the system stability condition compared to other response features.(2)The data-driven ANFIS based stability discrimination model establishes the if-then logical inference process between the input features(SPF and MAD)and the system stable/unstable status.It can detect the unstable condition in real-time and show higher accuracy of stability discrimination in a complex power grid compared to other AI stability discrimination methods.(3)The proposed stability analysis method maintains high accuracy even in the presence of changes in the power grid topology,showing good generalization performance.
作者
杨浩
伍柏臻
刘铖
孙正龙
蔡国伟
刘萌
Yang Hao;Wu Baizhen;Liu Cheng;Sun Zhenglong;Cai Guowei;Liu Meng(Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology Ministry of Education Northeast Electric Power University,Jilin 132012,China;State Grid Shandong Electric Power Research Institute,Jinan 250003,China)
出处
《电工技术学报》
EI
CSCD
北大核心
2024年第13期3943-3955,共13页
Transactions of China Electrotechnical Society
基金
国家重点研发计划资助项目(2021YFB2400800)。
关键词
暂态稳定性
响应驱动
特勒根定理
能量函数
复杂电网
自适应逻辑推理
Transient stability
response-driven
Tellegen’s theorem
transient energy function
complex power system
adaptive logical reasoning
