Theoretical approaches to detecting anomalies in meter readings in scientific literature
The article is devoted to the study of modern scientific approaches to detecting anomalies in meter readings in the context of the development of intelligent energy networks. The study’s relevance is due to the growing need for effective monitoring of electricity consumption, especially in the conte...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Zhytomyr Polytechnic State University
2025-07-01
|
| Series: | Технічна інженерія |
| Subjects: | |
| Online Access: | https://ten.ztu.edu.ua/article/view/334941 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The article is devoted to the study of modern scientific approaches to detecting anomalies in meter readings in the context of the development of intelligent energy networks. The study’s relevance is due to the growing need for effective monitoring of electricity consumption, especially in the context of martial law in Ukraine, when the energy system is under constant threat and requires the most efficient management of resources. The paper provides a comprehensive analysis of existing methods for detecting anomalies, which are classified into three main categories: point anomalies, which represent individual deviations from the typical consumption pattern; contextual anomalies, which depend on the specific temporal or spatial environment; and collective anomalies, which arise in groups of interrelated data. The study demonstrates the advantages of decentralised approaches to data processing, particularly the use of edge computing and fog computing technologies, which allow information to be analysed directly on metering devices or in their immediate vicinity. A comparative analysis of centralised and decentralised methods shows that hybrid systems, which combine the advantages of both approaches, demonstrate the best results in practical applications. Particular attention is paid to analysing the effectiveness of various machine learning algorithms for anomaly detection. Statistical methods based on probabilistic models are considered, which allow relatively accurate determination of normal limits provided sufficient historical data is available. Classification methods of supervised learning are analysed, which require pre-labelled data for model training but provide high accuracy in detecting known types of anomalies. The prospects for further development of anomaly detection technologies in the context of integrating IoT devices, technology development and implementing more complex deep learning algorithms are considered. |
|---|---|
| ISSN: | 2706-5847 2707-9619 |