Impact of distance measures in adaptive K-means clustering on load profiles and spatial patterns of distributed substations in Thailand
Authors:
Open Access
- 1 July 2025
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-20
Abstract
To address the challenges of increasing electricity demand and diverse consumption behavior, this study explored an adaptive K-means clustering approach for segmenting 24-h load profiles from 627 distributed substations of the Provincial Electricity Authority (PEA) in Thailand. Euclidean distance and Cosine similarity were applied as distance measures, with cluster numbers (K) ranging from 2 to 10. Clustering validity was evaluated using the Davies–Bouldin Index (DBI), Calinski–Harabasz Index (CHI), and Silhouette Coefficient (SC), alongside three proposed metrics: Averaged Standard Deviation (ASD), Standard Deviation of Derived Slope (SDDS), and Absolute Different Area (ADA). Euclidean distance was found to be more effective in clustering load profiles based on the magnitude of electricity consumption, while Cosine similarity better captured the shape and temporal patterns of usage, as supported by the proposed metrics. Optimal clustering for the distributed substations of PEA was achieved with K equal to 3 or 4, balancing simplicity and detail. The spatial distribution of substation clusters across different regions in Thailand revealed distinct energy consumption patterns linked to customer sectors. These findings provide valuable insights for electricity management strategies, distribution grid infrastructure planning, and future energy policy development in Thailand.Keywords
Funding Information
- Global and Frontier Research University Fund at Naresuan University (R2567C002)
- Naresuan University (NU) and the National Science, Research, and Innovation (R2568B093)
This publication has 48 references indexed in Scilit:
- Performance evaluation method for different clustering techniquesComputers and Electrical Engineering, 2025
- A machine learning-based framework for clustering residential electricity load profiles to enhance demand response programsApplied Energy, 2024
- Effects of similarity/distance metrics on k-means algorithm with respect to its applications in IoT and multimedia: a reviewMultimedia Tools and Applications, 2021
- Multi-view document clustering based on geometrical similarity measurementInternational Journal of Machine Learning and Cybernetics, 2021
- Recognition and classification of typical load profiles in buildings with non-intrusive learning approachApplied Energy, 2019
- Balancing the trade-off between accuracy and interpretability in software defect predictionEmpirical Software Engineering, 2018
- k-means based load estimation of domestic smart meter measurementsApplied Energy, 2017
- On the Surprising Behavior of Distance Metrics in High Dimensional Space2001
- Silhouettes: A graphical aid to the interpretation and validation of cluster analysisJournal of Computational and Applied Mathematics, 1987
- Principal component analysisChemometrics and Intelligent Laboratory Systems, 1987