Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Detection and discrimination of inrush current and faults in three-phase power transformers using signal processing and machine learning technique

Abdusalam, Othman 2023. Detection and discrimination of inrush current and faults in three-phase power transformers using signal processing and machine learning technique. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of Amended thesis.pdf] PDF - Accepted Post-Print Version
Restricted to Repository staff only until 6 December 2024 due to copyright restrictions.

Download (4MB)
[thumbnail of Cardiff University Electronic Publication Form] PDF (Cardiff University Electronic Publication Form) - Supplemental Material
Restricted to Repository staff only

Download (249kB)


Transformers are integral components of power systems, and their protection is critical to ensure the reliability and safety of the entire system. Internal faults in transformers are a major cause of failure, and thus, effective protection measures are essential to mitigate the risks associated with such faults. However, existing protection techniques are not infallible, and it is imperative to minimize damage to the transformer in the event of a fault, to reduce repair costs and prevent disruptions to the system. To address these issues, this thesis proposes three methods represented in the fourth study for transformer protection that are fast and highly effective in detecting both internal and external faults. The proposed techniques aim to overcome the challenge of inrush currents and saturation cases, which can lead to malfunctions in the protection system and lower the transformer's efficiency. The proposed protection technique is based on current signal processing and machine learning techniques, specifically, the use of alienation coefficients and machine learning classifiers based on practical and simulation tests. The proposed technique was successfully demonstrated using Simulink and simulations in the MATLAB software program and a laboratory model. The methods can identify and detect faults in phase A, with the same efficiency when applied to phases B and C. In the first study, the effectiveness of the alienation coefficient method in detecting common transformer faults, including cases of current overload, was demonstrated through a study where the method was applied to a transformer model created using MATLAB/SIMULINK. During steady state, it has been shown that the technique is efficient and fast at detecting faults as it detects faults between 2ms -3ms depending on the type of fault. In the second study, by using data from the laboratory, all possible faults were generated and imported to MATLAB for processing with the algorithm. The practical demonstration of the technique showed its effectiveness in detecting and eliminating internal faults within 3ms, depending on the strength of the fault current. The technique also proved to be capable of effectively differentiating between external and internal faults. Additionally, the test included examining current transformer (CT) saturation, which is a significant problem associated with protection schemes. In this study, the effectiveness of the alienation coefficient method was compared to a discrete Wavelet transform (DWT) method, both of which were applied to achieve the same objective. The proposed method was found to be faster than the DWT due to certain significant factors, with a fault being detected in just 3ms compared to the 5ms required by DWT. The third study proposes the use of three machine learning classifiers support vector machine (SVM), k-nearest neighbors (KNN) and artificial neural network (ANN), - for fault type detection in power transformers. The classifiers were trained on selected features and showed high accuracy in fault detection and classification. An optimization algorithm-based feature selection was performed to validate the effectiveness of the model, revealing that the proposed strategies with fewer statistical features provided more accurate results. Combining the Bees algorithm with an SVM classifier resulted in maximal accuracy of 96% in fault classification, with high accuracy observed in evaluation-based testing. The fourth study proposes a novel technique for fault classification in a three-phase power transformer using SVM. The method uses two classifiers, SVM1 for fault detection and SVM2 for identifying internal and external faults. The technique achieves fault identification and detection in less than a quarter cycle of the phase current of phase A, with the same technique applied to phases B and C. The classification of faults is not affected by the type or location of the fault, and the SVM method is shown to be more accurate and faster than the conventional ANN method. The proposed technique is effective in detecting different types of faults with high accuracy under different operating conditions and provides an efficient and reliable approach for fault classification in power transformers.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Engineering
Uncontrolled Keywords: 1). external fault alienation strategy 2). internal fault 3). MATLAB software 4). LabVIEW software 5). inrush current
Date of First Compliant Deposit: 7 December 2023
Last Modified: 07 Dec 2023 13:55

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics