An efficient computational method for network analysis using clustering of electric vehicle charging pattern in parking

Abstract

Due to the increasing penetration of electric vehicles (EVs) in the transportation fleet, long-term network stability analyses have faced serious challenges due to the large amount of data and computational complexity. This issue is due to the need to examine many daily scenarios to understand the long-term impacts of vehicle charging on the network. This process is considered very time-consuming and impractical. In response to this need, this paper introduces a method for clustering long-term scenarios in EV charging planning, which aims to reduce the computational burden in network stability studies. The proposed approach includes the (main) steps of collecting EV data from parking lots (PLs), extracting probability density functions for key parameters such as entry/exit times and charging rates, and then generating synthetic data for many scenarios. In the next step, the EV data tables are converted into feature vectors and clustered using the K-means and Fuzzy C-means clustering algorithms. One of the key innovations of this research is the provision of a robust validation framework obtained by comparing the results of two clustering paths: the first path based on EV data tables (EVDT) and the second path based on daily load curves (DLC) generated by the charging scheduling algorithm. Simulation results performed on real data show that the proposed method can correctly identify four distinct behavioral patterns. This means that the best number of clusters is determined to be four. Furthermore, the results obtained from the two clustering paths, with more than 90 % similarity, confirm the reliability and efficiency of the method in extracting representative scenarios. The findings of the paper indicate that this method could significantly reduce the computational volume associated with long-term network studies while maintaining the accuracy and comprehensiveness of the analysis.