The Application of FFT Analysis on the Feeder Current Based on the Smooth Function in MATLAB

Yujie Chang, Huawei Li, Songwei Huang, Jinghan He, Shaobing Yang, Qiang WangAdvances 2015 · DOI

Abstract

Fast Fourier transform (FFT) is an important tool for harmonic analysis in power systems and a method of FFT analysis of the signal is very important for power systems. For finding and forming a basic way to FFT analysis of feeder current, the paper embarks from the principle of Fourier transform and analyzes the principle and algorithm of the related metrics such as spectrum and signal-to-noise ratio (SNR). Then it uses the smooth function which is based on the moving average filter to filter the feeder current, which is used to prove the feasibility of this method. Meanwhile its result which we can get the data from figures and tables in the paper also shows that the smooth function as a low-pass filter can effectively filter out noise and keep the feeder current fundamental wave using the way that changes span values and method, but it will cause the base wave distortion.

Introduction

In modern electrified railways, high-power locomotives act as non-linear loads that inject significant harmonics into the power grid. These harmonics cause electrical equipment to overheat, degrade power quality, and threaten the stability of the entire power system. To monitor and protect these systems, engineers need to accurately analyze the "feeder current"—the electricity supplied to the overhead lines. However, real-world feeder current signals are often buried in noise. This study explores how to effectively clean up these signals using MATLAB’s smooth function and Fast Fourier Transform (FFT) analysis, identifying the critical trade-off between noise suppression and signal distortion.

Methods and Data

The study utilizes two primary data sources: a synthetic test signal for algorithm validation and real-world field data captured from a traction substation. The raw feeder current signal contains the 50Hz fundamental plus various odd harmonics (150Hz, 250Hz, etc.).

Time-domain representation of the raw feeder current signal.

The researchers combine classic signal processing theory with practical MATLAB implementation:

  1. FFT Analysis: Moving the signal from the time domain to the frequency domain to identify the fundamental 50Hz wave and its harmonics.
  2. SNR Estimation: Calculating the Signal-to-Noise Ratio by comparing the energy at the target frequency point against the rest of the spectrum.
  3. MATLAB smooth Function: Testing six different filtering algorithms (Moving, Lowess, Loess, Sgolay, Rlowess, Rloess) while systematically varying the span (the number of data points included in each local average).

Results and the "Span Paradox"

The raw feeder current had an initial SNR of approximately 9.8168. By applying the 'moving' method with an optimized span of 57, the SNR improved to 18.5866, nearly doubling the signal clarity. Surprisingly, the simplest method—the moving average—outperformed more complex polynomial regression methods like loess or sgolay for this specific application.

However, a critical challenge emerged: the "Span Paradox." Increasing the span of the filter is a double-edged sword. While a larger span does a better job of smoothing out high-frequency noise, it simultaneously reduces the amplitude of the fundamental 50Hz wave. This is clearly visible in the experimental data: with a small span of 17, the 50Hz amplitude was 23.29; however, when the span was increased to 97 to maximize smoothing, the amplitude dropped significantly to 15.10.

Comparison of filtered signals across different span values.

Conclusion

The research demonstrates that the MATLAB smooth function is a feasible tool for improving the SNR of feeder current signals. While the moving average filter effectively suppresses noise, it introduces "base wave distortion" by reducing the peak amplitude of the fundamental frequency. Future work could explore adaptive filtering methods or hybrid approaches, such as combining the smooth function with advanced wavelet transforms, to better preserve the fundamental wave while maintaining high noise suppression.

Summary updated

Abstract

Fast Fourier transform(FFT) is an important tool for harmonic analysis in power systems and a method of FFT analysis of the signal is very important for power systems.For finding and forming a basic way to FFT analysis of feeder current, the paper embarks from the principle of Fourier transform and analyzes the principle and algorithm of the related metrics such as spectrum and signal-to-noise ratio(SNR).Then it uses the smooth function which is based on the moving average filter to filter the feeder current, which is used to prove the feasibility of this method.Meanwhile its result which we can get the data from figures and tables in the paper also shows that the smooth function as a low-pass filter can effectively filter out noise and keep the feeder current fundamental wave using the way that changes span values and method, but it will cause the base wave distortion.
BibTeX
@inproceedings{chang2015application,
  author = {Yujie Chang and Huawei Li and Songwei Huang and Jinghan He and Shaobing Yang and Qiang Wang},
  title = {{The Application of FFT Analysis on the Feeder Current Based on the Smooth Function in MATLAB}},
  booktitle = {Advances},
  year = {2015},
  doi = {10.2991/ap3er-15.2015.102},
}
Figure