Voltage disturbance detection and classification in distribution networks using simulated events : A data-driven framework using realistic load profiles and machine learning
Zaid, Zaid (2025)
2025
Sähkötekniikan DI-ohjelma - Master's Programme in Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2025-12-13
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-2025111410618
https://urn.fi/URN:NBN:fi:tuni-2025111410618
Tiivistelmä
The increasing penetration of distributed energy resources (DERs) such as photovoltaic (PV) systems, electric vehicles (EVs), and heat pumps has made modern low-voltage (LV) distribution networks more dynamic and prone to voltage disturbances. These events, arising from rapid fluctuations in load or generation, can cause voltage deviations that complicate voltage control coordination. This thesis presents a data-driven, simulation-based framework for detecting and classifying voltage disturbances in distribution networks, with the aim of supporting coordinated voltage control (CVC) strategies.
A detailed interconnected HV–MV–LV grid model was developed in MATLAB to simulate disturbance scenarios under realistic time-series load profiles. Synthetic voltage and current data were generated by injecting diverse disturbance types—such as load surges and upstream voltage sags—under varying network conditions. Key monitoring indicators, including voltage dips, current magnitudes, and statistical variations of node voltages, were extracted as features for classification. Multiple machine learning algorithms, including decision trees, support vector machines (SVM), k-nearest neighbors (k-NN), and ensemble methods, were trained and evaluated on their ability to distinguish both the type and origin of disturbances (HV, MV, or LV).
Results demonstrate that machine learning models can reliably identify the disturbance type and origin with high accuracy when trained on sufficiently resolved monitoring data. The analysis further highlights how data resolution, feature selection, and disturbance characteristics influence detection reliability. The proposed framework establishes a foundation for integrating intelligent disturbance detection and classification into real-time coordinated voltage control (CVC) systems, thereby improving grid resilience and operational efficiency.
A detailed interconnected HV–MV–LV grid model was developed in MATLAB to simulate disturbance scenarios under realistic time-series load profiles. Synthetic voltage and current data were generated by injecting diverse disturbance types—such as load surges and upstream voltage sags—under varying network conditions. Key monitoring indicators, including voltage dips, current magnitudes, and statistical variations of node voltages, were extracted as features for classification. Multiple machine learning algorithms, including decision trees, support vector machines (SVM), k-nearest neighbors (k-NN), and ensemble methods, were trained and evaluated on their ability to distinguish both the type and origin of disturbances (HV, MV, or LV).
Results demonstrate that machine learning models can reliably identify the disturbance type and origin with high accuracy when trained on sufficiently resolved monitoring data. The analysis further highlights how data resolution, feature selection, and disturbance characteristics influence detection reliability. The proposed framework establishes a foundation for integrating intelligent disturbance detection and classification into real-time coordinated voltage control (CVC) systems, thereby improving grid resilience and operational efficiency.