Sizing and Characterization of Load Curves of Distribution Transformers Using Clustering and Predictive Machine Learning Models

Pedro Torres-Bermeo, Kevin López-Eugenio, Carolina Del-Valle-Soto, Guillermo Palacios-Navarro and José Varela-Aldás ()
Additional contact information
Pedro Torres-Bermeo: Centro de Investigación MIST, Facultad de Ingenierías, Maestría en Big Data y Ciencia de Datos, Universidad Tecnológica Indoamérica, Ambato 180103, Ecuador
Kevin López-Eugenio: Centro de Investigación MIST, Facultad de Ingenierías, Maestría en Big Data y Ciencia de Datos, Universidad Tecnológica Indoamérica, Ambato 180103, Ecuador
Carolina Del-Valle-Soto: Facultad de Ingeniería, Universidad Panamericana, Álvaro del Portillo 49, Zapopan 45010, Mexico
Guillermo Palacios-Navarro: Department of Electronic Engineering and Communications, University of Zaragoza, 44003 Teruel, Spain
José Varela-Aldás: Centro de Investigación MIST, Facultad de Ingenierías, Maestría en Big Data y Ciencia de Datos, Universidad Tecnológica Indoamérica, Ambato 180103, Ecuador

Energies, 2025, vol. 18, issue 7, 1-24

Abstract: The efficient sizing and characterization of the load curves of distribution transformers are crucial challenges for electric utilities, especially given the increasing variability of demand, driven by emerging loads such as electric vehicles. This study applies clustering techniques and predictive models to analyze and predict the behavior of transformer demand, optimize utilization factors, and improve infrastructure planning. Three clustering algorithms were evaluated, K-shape, DBSCAN, and DTW with K-means, to determine which one best characterizes the load curves of transformers. The results show that DTW with K-means provides the best segmentation, with a cross-correlation similarity of 0.9552 and a temporal consistency index of 0.9642. For predictive modeling, supervised algorithms were tested, where Random Forest achieved the highest accuracy in predicting the corresponding load curve type for each transformer (0.78), and the SVR model provided the best performance in predicting the maximum load, explaining 90% of the load variability (R 2 = 0.90). The models were applied to 16,696 transformers in the Ecuadorian electrical sector, validating the load prediction with an accuracy of 98.55%. Additionally, the optimized assignment of the transformers’ nominal power reduced installed capacity by 39.27%, increasing the transformers’ utilization factor from 31.79% to 52.35%. These findings highlight the value of data-driven approaches for optimizing electrical distribution systems.

Keywords: machine learning; clustering; transformer load characterization; loadability; predictive modeling; DTW with K-means; Support Vector Machines; Random Forest (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.mdpi.com/1996-1073/18/7/1832/pdf (application/pdf)
https://www.mdpi.com/1996-1073/18/7/1832/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:18:y:2025:i:7:p:1832-:d:1628262

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().