融合关键机组信息与相平面图像的暂态稳定评估方法

Transient stability assessment base on fusion of critical generator information and state plane images

针对传统基于电力系统时序数据的机器学习方法对系统深层耦合信息挖掘不充分的问题,将失稳模式作为先验知识引入机器学习,提出一种融合系统关键机组信息的暂态稳定性评估方法.该方法通过潮流追踪原理对线路计算各发电机潮流贡献度,得出系统关键机组权重.根据图像形态学原理,对相平面轨迹图像依照关键机组权重进行特征增强.在IEEE-39节点和IEEE-145节点系统下的仿真结果表明,所提方法较传统评估方法具有更好的评估性能,所构建的相平面图像样本较传统时序图像样本拥有更小的占用空间和更优的分类性能.

Aiming at the problem that traditional machine learning methods based on power system time-series data are inadequate in mining the deep coupling information of the system,this paper introduces the physical information of instability mode as a priori knowledge into machine learning,and proposes a transient stability assessment method that incorporates the information of critical generators.The method calculates the contribution of each generator's power flow to the line through the power flow tracing to derive the system critical generator weights.According to the morphology,the state plane trajectory images are feature-enhanced according to the key generator weights.Simulation results under IEEE 39-bus system and IEEE 145-bus system show that the proposed method has better evaluation performance than the traditional evaluation method;the constructed state plane image samples have smaller occupation space and better classification performance than the traditional time series image samples.