Efficient Data Reduction in Multisource Hybrid Electrical Systems Using Frequency-Domain Decimation Techniques

Abstract

This work addresses the challenge of reducing data size without compromising analytical accuracy in simulations of a hybrid multisource electrical system. The system generates large volumes of time-series data—voltage, current, and power waveforms whose storage and processing can become computationally expensive. To mitigate this, we propose a reduction strategy based on the Fast Fourier Transform (FFT). A decimation coefficient is computed for each variable in the frequency domain, and the signals are then reconstructed using the Inverse FFT (IFFT). Simulation results show that the method achieves a substantial decrease in data volume while preserving the essential features of the original signals, ensuring the reliability of subsequent analyses and machine-learning-based fault diagnostics.