Research on Distance Preservation Ability of Fault Location Features for Data-driven Dynamic Security Assessment
Zhe Ren, Hang Qi, Changgang Li, Hanbing Qu, Junjun Yang, Wei Xu · 2022
In the feature expression of fault location in the data-driven dynamic security assessment, Euclidean distance is generally used to measure the distance between features. However, Euclidean distance cannot accurately reflect the electrical distance relationship between the described fault locations, which limits the further improvement of the accuracy of data-driven dynamic security assessment. This paper proposes the concept of strict distance preservation of fault location features. Based on the fault location feature expression based on the electrical coordinate system, the mathematical definition of strict distance preservation of electrical coordinates is constructed. Based on the distance metric of p-norm, the conjecture that Chebyshev distance is the only distance metric to satisfy strict distance preservation is proposed and proved. Furthermore, the relationship between the distance preservation ability of electrical coordinates and p-norm is analyzed under the Chebyshev distance metric. Taking the transient stability crucial clearing time as an example, the contribution of the preservation ability of electrical coordinates to the accuracy of dynamic security assessment is verified based on K-nearest neighbor algorithm.