A new method of power system fault recording based on compressed sensing

Abstract

Wavelet transform is usually used to deal with the numerous data in fault recording of power systems, but it still needs to sample the signal with a high frequency first and then compress the data obtained. In order to solve the problem, this paper proposes a new fault recording method based on compressed sensing theory. The method can break the limit of Nyquist sampling theorem and acquire the fault signal with a much lower sampling frequency. It presents the theoretical framework of compressed sensing, analyzes the characteristics of the faulty waveform, and proposes a new sparsity adaptive and compressive sampling matching pursuit algorithm to reconstruct the fault signal. Finally, simulation results in MATLAB show that the proposed method performs well at different compression ratios in terms of the norm mean-square error, signal-to-noise ratio, and waveform similarity.

State estimation and fault reconstruction with integral measurements under partially decoupled disturbances

“…Consequently, the research on the fault diagnosis (FD) problem has been gaining momentum and considerable effort has been devoted to the detection of occurrence of faults, the identification of the position of faults as well as the determination of the amplitude of faults. So far, the FD problem has been extensively studied in many practical systems including induction motors [1], robotic systems [2], wind turbines [3], power systems [4], planetary gearboxes [5], three‐phase inverters [6], flight control systems [7, 8] and so forth. Among existing FD approaches, model‐based FD methods have been widely adopted and a great number of results have been available in the literature on model‐based FD issues, see, e.g.…”