Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data
Accurate electrical phase identification (PI) is essential for efficient grid management, yet existing research predominantly focuses on high-frequency smart meter data, not adequately addressing phase identification with low sampling rates using energy consumption data. This study addresses this ga...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-12-01
|
| Series: | Energies |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1996-1073/18/1/128 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841549249954709504 |
|---|---|
| author | Dilan C. Hangawatta Ameen Gargoom Abbas Z. Kouzani |
| author_facet | Dilan C. Hangawatta Ameen Gargoom Abbas Z. Kouzani |
| author_sort | Dilan C. Hangawatta |
| collection | DOAJ |
| description | Accurate electrical phase identification (PI) is essential for efficient grid management, yet existing research predominantly focuses on high-frequency smart meter data, not adequately addressing phase identification with low sampling rates using energy consumption data. This study addresses this gap by proposing a novel method that employs a fully connected neural network (FCNN) to predict household phases from energy consumption data. The research utilizes the IEEE European Low Voltage Testing Feeder dataset, which includes one-minute energy consumption readings for 55 households over a full day. The methodology involves data cleaning, preprocessing, and feature extraction through recursive feature elimination (RFE), along with splitting the data into training and testing sets. To enhance performance, training data are augmented using a generative adversarial network (GAN), achieving an accuracy of 91.81% via 10-fold cross-validation. Additional experiments assess the model’s performance across extended sampling intervals of 5, 10, 15, and 30 min. The proposed model demonstrates superior performance compared to existing classification, clustering, and AI methods, highlighting its robustness and adaptability to varying sampling durations and providing valuable insights for improving grid management strategies. |
| format | Article |
| id | doaj-art-1d595922e0874630ab6aa0fae6f3c329 |
| institution | Kabale University |
| issn | 1996-1073 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Energies |
| spelling | doaj-art-1d595922e0874630ab6aa0fae6f3c3292025-01-10T13:17:10ZengMDPI AGEnergies1996-10732024-12-0118112810.3390/en18010128Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer DataDilan C. Hangawatta0Ameen Gargoom1Abbas Z. Kouzani2School of Engineering, Deakin University, Geelong, VIC 3216, AustraliaSchool of Engineering, Deakin University, Geelong, VIC 3216, AustraliaSchool of Engineering, Deakin University, Geelong, VIC 3216, AustraliaAccurate electrical phase identification (PI) is essential for efficient grid management, yet existing research predominantly focuses on high-frequency smart meter data, not adequately addressing phase identification with low sampling rates using energy consumption data. This study addresses this gap by proposing a novel method that employs a fully connected neural network (FCNN) to predict household phases from energy consumption data. The research utilizes the IEEE European Low Voltage Testing Feeder dataset, which includes one-minute energy consumption readings for 55 households over a full day. The methodology involves data cleaning, preprocessing, and feature extraction through recursive feature elimination (RFE), along with splitting the data into training and testing sets. To enhance performance, training data are augmented using a generative adversarial network (GAN), achieving an accuracy of 91.81% via 10-fold cross-validation. Additional experiments assess the model’s performance across extended sampling intervals of 5, 10, 15, and 30 min. The proposed model demonstrates superior performance compared to existing classification, clustering, and AI methods, highlighting its robustness and adaptability to varying sampling durations and providing valuable insights for improving grid management strategies.https://www.mdpi.com/1996-1073/18/1/128phase identificationlow sampling rateenergy consumption datafully connected neural networkrecursive feature eliminationgenerative adversarial network |
| spellingShingle | Dilan C. Hangawatta Ameen Gargoom Abbas Z. Kouzani Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data Energies phase identification low sampling rate energy consumption data fully connected neural network recursive feature elimination generative adversarial network |
| title | Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data |
| title_full | Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data |
| title_fullStr | Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data |
| title_full_unstemmed | Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data |
| title_short | Machine-Learning-Driven Identification of Electrical Phases in Low-Sampling-Rate Consumer Data |
| title_sort | machine learning driven identification of electrical phases in low sampling rate consumer data |
| topic | phase identification low sampling rate energy consumption data fully connected neural network recursive feature elimination generative adversarial network |
| url | https://www.mdpi.com/1996-1073/18/1/128 |
| work_keys_str_mv | AT dilanchangawatta machinelearningdrivenidentificationofelectricalphasesinlowsamplingrateconsumerdata AT ameengargoom machinelearningdrivenidentificationofelectricalphasesinlowsamplingrateconsumerdata AT abbaszkouzani machinelearningdrivenidentificationofelectricalphasesinlowsamplingrateconsumerdata |