A comprehensive guide to selecting suitable wavelet decomposition level and functions in discrete wavelet transform for fault detection in distribution networks

Esraa M Shalby; Almoataz Y Abdelaziz; Eman S Ahmed; Basem Abd-Elhamed Rashad

doi:10.1038/s41598-024-82025-2

A comprehensive guide to selecting suitable wavelet decomposition level and functions in discrete wavelet transform for fault detection in distribution networks

Sci Rep. 2025 Jan 7;15(1):1160. doi: 10.1038/s41598-024-82025-2.

Authors

Esraa M Shalby^{1

2}, Almoataz Y Abdelaziz^{3

4}, Eman S Ahmed⁵, Basem Abd-Elhamed Rashad⁶

Affiliations

¹ Faculty of Engineering, Ain Shams University, Cairo, 11517, Egypt. [email protected].
² Department of Electrical Power and Machines Engineering, The Higher Institute of Engineering at El- Shorouk City, El-Shorouk Academy, Cairo, 11837, Egypt. [email protected].
³ Faculty of Engineering, Ain Shams University, Cairo, 11517, Egypt.
⁴ Faculty of Engineering and Technology, Future University in Egypt, Cairo, 11835, Egypt.
⁵ Department of Electrical Engineering, Faculty of Engineering, Kafrelsheikh University, Kafrelsheikh, Egypt.
⁶ Department of Electrical Power and Machines Engineering, The Higher Institute of Engineering at El- Shorouk City, El-Shorouk Academy, Cairo, 11837, Egypt.

Abstract

The paper presents a comprehensive analysis of the IEEE-16 bus system under different operating conditions. It discusses the selection of suitable decomposition level and wavelet function for analyzing non-stationary signals to enhance power distribution network fault detection. MATLAB/Simulink is used to simulate the system, and transient fault current signals are processed with the MATLAB Wavelet Toolbox. The optimal decomposition level is determined by energy concentration, with the highest energy found in scales D9 (b4), D8 (b5), and D7 (b6), and D8 having the most concentration. Using MATLAB classifier learner, the article evaluates seven common mother wavelets with 53 wavelet functions, and sym3 is found to be the most efficient wavelet function in terms of training time, prediction speed, and accuracy of SVM classifiers. All fault types both symmetrical/unsymmetrical types, and various normal transient conditions such as load/capacitor/DG switching are detected/discriminated with nearly 100% accuracy at the midpoint of line 6-7 with various fault conditions, inception angles (0, 30, 45, 60, 90 and 120°) and a fault resistance of (5,10, 15, and 20 ohms). Additionally, 9 MW wind Farm is integrated at busbar 10, and various fault scenarios are simulated to assess system performance with 100% Accuracy.

Keywords: Decomposition level selection; Discrete wavelet transform (DWT); Distribution networks; Fault detection; Feature extraction; MATLAB classifier learner; MATLAB wavelets toolbox; Mother wavelets selection; Support vector machine.