Adaptive real-time DWT-based method for arc fault detection

The real-time detection of arc faults is crucial in many industrial applications. Many research works based on the use of Discrete Wavelet Transform (DWT) have been done in the last years. There are three parameters that have a decisive influence on the performance of DWT for arc fault detection: the sampling frequency, the mother wavelet and the level of decomposition. For a specific load, an optimal combination of these three parameters must be found. Most research works use one particular combination of these parameters for a specific load (mainly resistive). This study shows that there is no unique combination for all types of loads. In this paper, we propose an adaptive system which can update automatically these three parameters for each type of load. The first phase of the proposed method is validated with MATLAB by considering all possible combinations: 10 different mother wavelets, 10 levels of decomposition and 10 values of sampling frequency. The result of this first phase is an optimal combination of these three parameters for a given load. In the second phase, we implement the computed optimal solution from the first phase on a hardware platform. Thus, we propose an adaptive FIR-based reconfigurable architecture which can host every optimal combination. The FPGA architecture, tested by using Hardware-in-the-loop approach, shows quite good performances for all tested types of load.