Abstract
Conference Title: 2017 IEEE Energy Conversion Congress and Exposition (ECCE) Conference Start Date: 2017, Oct. 1 Conference End Date: 2017, Oct. 5 Conference Location: Cincinnati, OH, USA Dc arc faults may result in serious damage to both stationary and mobile dc power systems. This paper presents a new detection method for dc arcs using Hurst exponents. Compared to other methods such as the Fast Fourier Transform (FFT) and the Wavelet Transform, the proposed method is adaptive and is more immune to power electronics switching harmonics. A test setup to generate dc arcs is built and the Hurst exponent algorithm is applied. The algorithm is studied under different load conditions and noisy environments. The Hurst exponent method is shown to be effective in detecting dc arcs within hundreds of milliseconds and bot miss-triggered by step load changes, switching harmonics, and load profiles.