An embedded system for AC series arc detection by inter-period correlations of current

This paper presents a method for detecting series arcs for a 230 V AC-50 Hz residential installation, its implementation in an embedded circuit and its performance under real conditions. First, we evaluated the inter-period autocorrelation coefficient r of the line current. The algebraic estimation of the derivative of the signal r was then used to obtain an indicator of arc detection. r was calculated on the current at a sampling rate frequency of 5 kHz and the 50 Hz component was attenuated by an high pass (HP) filter. Tests carried out with domestic appliances with different power loads (resistive load, lamp, vacuum cleaner, dimmer etc) show a value for r which is very stable and close to 1 under normal conditions. The coefficient r weakens in the presence of an arc (initiated by a carbon path) and is no longer constant. The complete algorithm was implemented in real time on a 32-bit ARM Cortex-M3 microcontroller with a 12-bit ADC. The measuring current, which ranges from 0 to more than 20 A, was provided by a passive toroidal inductive probe. Four typical household loads were tested. The results obtained show that detection is operative and very fast (10 ms) with currents which exceed 400 mA.