Real-time detection of high-impedance faults in a distribution circuit is described. The real-time detection of high-impedance faults includes two steps. First, adaptive soft denoising is employed to perform a filtering process on a healthy dataset, and to determine a threshold. This reduces the rate of false alarms. Second, faulty datasets are prefiltered via adaptive soft denoising, then the denoised signals are processed via discrete wavelet transform to perform high-impedance fault detection using the threshold.