[Paper Review] A matrix approach to detect temporal behavioral patterns at electric vehicle charging stations

Milan Straka, Lucia Piatriková+2 more|arXiv (Cornell University)|Feb 18, 2021

Electric Vehicles and Infrastructure6 references4 citations

TL;DR

This paper proposes a matrix-based approach to detect temporal charging patterns at electric vehicle (EV) charging stations using two methods: a rule-based system for predefined patterns and hierarchical clustering with a modified l-p norm dissimilarity measure for automated pattern discovery. The clustering approach successfully identified 10 distinct behavioral clusters, including morning, evening, and mixed charging patterns, with optimal performance at parameters o=1, p=2/3.

ABSTRACT

Based on the electric vehicle (EV) arrival times and the duration of EV connection to the charging station, we identify charging patterns and derive groups of charging stations with similar charging patterns applying two approaches. The ruled based approach derives the charging patterns by specifying a set of time intervals and a threshold value. In the second approach, we combine the modified l-p norm (as a matrix dissimilarity measure) with hierarchical clustering and apply them to automatically identify charging patterns and groups of charging stations associated with such patterns. A dataset collected in a large network of public charging stations is used to test both approaches. Using both methods, we derived charging patterns. The first, rule-based approach, performed well at deriving predefined patterns and the latter, hierarchical clustering, showed the capability of delivering unexpected charging patterns.

Motivation & Objective

• To identify temporal behavioral patterns in EV charging station usage based on arrival times and connection durations.
• To compare rule-based categorization with unsupervised clustering for detecting charging patterns.
• To evaluate the effectiveness of a modified l-p norm as a matrix dissimilarity measure in hierarchical clustering for pattern discovery.
• To group charging stations with similar temporal usage patterns for improved prediction and system optimization.
• To explore the potential of clustering for enhancing smart charging and power grid stability.

Proposed method

• Represent EV charging data as a charging matrix with hourly arrival times and connection durations.
• Apply a rule-based approach using predefined time intervals and threshold values to classify stations into known charging patterns.
• Use a modified l-p norm as a matrix dissimilarity measure to quantify differences between charging matrices.
• Combine the dissimilarity measure with agglomerative hierarchical clustering using complete-linkage to group stations.
• Optimize clustering by testing parameter combinations (o, p) in the modified l-p norm and evaluating cluster distribution and interpretability.
• Visualize results using heatmaps of normalized element-wise sums of charging matrices within each cluster.

Experimental results

Research questions

• RQ1 Can rule-based methods effectively identify predefined temporal charging patterns in EV charging stations?
• RQ2 Can hierarchical clustering with a custom matrix dissimilarity measure automatically detect unexpected or novel charging patterns?
• RQ3 Which parameter values (o, p) in the modified l-p norm yield the most interpretable and balanced clustering results?
• RQ4 How do the identified clusters of charging stations differ in terms of temporal behavior (e.g., morning vs. evening charging)?
• RQ5 To what extent can clustering of charging stations improve the accuracy of EV charging behavior prediction models?

Key findings

• The rule-based approach successfully identified and grouped charging stations according to predefined patterns, such as morning-only or overnight charging.
• Hierarchical clustering with the modified l-p norm detected 10 distinct clusters, including clusters with predominantly short, morning-long, or evening-long charging.
• The optimal parameter combination for clustering was o=1, p=2/3, which produced the most balanced and interpretable cluster distribution.
• Cluster 1 showed mostly short charging durations, while Clusters 2, 5, and 10 exhibited high morning charging activity, and Clusters 3 and 8 showed strong evening charging patterns.
• The remaining clusters (4, 6, 7, 9) displayed a uniform distribution of charging across multiple time periods, indicating mixed usage patterns.
• The study demonstrates that matrix-based clustering can uncover meaningful, non-intuitive temporal patterns beyond pre-defined rules, enhancing smart charging strategies.