Towards Smarter Energy Management: Clustering Electric Vehicle Charging and Trading Patterns
2025 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE credits
Student thesis
Abstract [en]
As more electric vehicles (EVs) take to the roads, understanding how they are charged is essential for better energy management and a more flexible power grid. This thesis looks at how AI and data mining, especially clustering techniques, can be used to study EV charging habits in private, commercial, and industrial settings. The study compares two time-series clustering approaches—one that preserves the natural flow of charging data (shape-based clustering using Dynamic Time Warping- DTW) and another that simplif ies the data into key features for analysis (feature-based clustering using K-Means and hierarchical clustering). The goal is to identify patterns in charging habits and key factors influencing them, such as state of charge (SOC), charging duration, and weather conditions.
The research is based on real EV charging data from Swedish energy companies like Eskilstuna Energi och Milj¨o (ESEM) and Uddevalla Energi. Since the datasets are large and complex, data cleaning and preprocessing were essential to handle missing values, remove outliers, and ensure consistency. Exploratory Data Analysis (EDA) was then used to uncover trends and correlations in charging behavior.
Clustering based on shape proved to be quite demanding in terms of computing resources, and it didn’t always reflect the time-related patterns in the data as clearly as expected. On the other hand, using extracted features made the process more efficient and much easier to interpret. Out of all the methods I explored, Agglomerative Clustering stood out by forming clear and meaningful groups. These results could be especially helpful for grid operators and energy providers who are working to improve how electric vehicle charging is managed and integrated into the energy system.
Place, publisher, year, edition, pages
2025. , p. 65
Keywords [en]
Clustering, Time-Series Analysis, Electric Vehicles, Charging Behavior, Data Mining, Smart Grid, AI, Renewable Energy Integration
National Category
Energy Engineering
Identifiers
URN: urn:nbn:se:mdh:diva-72654OAI: oai:DiVA.org:mdh-72654DiVA, id: diva2:1981319
External cooperation
Eskilstuna Energi och Miljö (ESEM), Uddevalla Energi
Subject / course
Energy Engineering
Supervisors
Examiners
2025-07-032025-07-032025-10-10Bibliographically approved