High-impedance faults (HIFs) pose the greatest challenge for distribution system protection, especially for microgrids and distribution networks with distributed generators (DGs) that have flexible operation modes. This paper analyzes the faulty features of HIFs and proposes a HIF detection method that uses empirical wavelet transform (EWT) and differential faulty energy. The proposed method is as follows. First, the various time-frequency components are obtained by utilizing the EWT to decompose the differential faulty energy and adaptively select the feature component with the largest permutation entropy. Second, the permutation variance index is constructed based on the sample point number and feature component energy, and then it is employed to detect HIFs. Finally, low voltage microgrid simulation tests, medium voltage distribution system integrated by DG simulation tests, and field tests show that the proposed method can correctly distinguish HIFs from normal disturbances, including operation mode switches, load switches, capacitor switches, and DG switches. The advantages of the proposed method are also elaborated in detail, from signal preprocessing and feature extraction.

In the event of a single line to ground (SLG) fault in an ineffectively grounded distribution network, the faulted current is weak and the existing devices cannot accurately judge the faulted feeder. In this paper, we proposed algorithms that combine CEEMDAN and Hilbert-transform to construct a multi-criteria comprehensive voting method. First, CEEMDAN algorithm is used to decompose the zero-sequence current to obtain the IMF1 (the first intrinsic mode function) component and the Hilbert transform is used to calculate the instantaneous amplitude and instantaneous phase. Then, according to the three largest instantaneous amplitudes information, we constructed the CIP, CIE relative entropy and CICP criteria from the phase, energy and polarity, respectively. Last, we proposed a comprehensive voting method, which is specifically shown as follows: when two or more criteria show that one feeder or the bus has a SLG fault, it is voted that the feeder or the bus has a SLG fault. In contrast, if the judgment results of the three criteria are inconsistent, then we would return to recalculation and then vote. Compared with existing method, simulation tests and field experiments show that the method proposed in this paper has higher accuracy and a faster calculation speed.

When a single-phase-to-ground (SPG) fault occurs in distribution networks, the existing faulty feeder detection techniques do not present reliable detection results. This paper introduces a novel faulty feeder detection method based on transient energy and cosine similarity. First, the fault characteristics of transient zero-sequence current (TZSC) are analyzed. It is found that the fault transient characteristics of transient components decrease as the grounding resistance increases. Second, to fully maintain the fault transient features, a fundamental frequency component (FFC) shifting method is introduced to remove FFCs. Next, the transient currents without containing FFCs are utilized to generate two fault indicators: transient energy within the first half power cycle and cosine similarity after the first half power cycle. After that step, to effectively merge the two fault indicators, the Laplace distribution is utilized. The fault possibility of each feeder can be quantitatively estimated by calculating the cumulative density function (CDF). Finally, the feeder with the maximum fault possibility is determined as the faulty feeder. The effectiveness and applicability of the proposed detection method are verified in the radial distribution network and the modified IEEE 34-bus distribution network. The results of the field test also indicate the feasibility of the proposed SPG detection method.

The problem of high impedance fault (HIF) detection in resonant grounding distribution systems has not been well solved. In practical applications, the existing detection methods either have complex calculation processes and low accuracy or have a weak ability to adapt to complex fault conditions, which limits the adaptability of the existing methods. Based on a theoretical analysis, a novel faulty feeder detection method based on the inner product transformation of extreme values is proposed in this paper. First, to eliminate the interference of high-frequency components, the low-frequency characteristic modes of zero-sequence current of each feeder are extracted by the variational mode decomposition (VMD) algorithm. On this basis, the low-frequency intrinsic mode functions (IMFs) are processed by the first-order difference to obtain the current waveform after the difference. Second, the zero crossing point of the differential current waveform is obtained, and then the extreme value point of the original zero-sequence current is calibrated, including the maximum and minimum. Third, at the extreme value point, the comprehensive inner product values (CIPV) of each feeder are calculated. Then, when the values are positive, it is judged as a bus fault, when the value has at least one negative value, it is judged as a feeder fault, and the feeder with the minimum value is judged as a faulty feeder. A large number of simulations and field tests show that this method has better fault detection accuracy in practical applications and is suitable for strong noise interference and asynchronous sampling.

When a high-impedance fault (HIF) occurs in a distribution network, the fault current is very weak and can be easily confused with capacitor switching (CS), load switching (LS) or inrush current (IC). This can lead to an inaccurate judgment of current protection. Furthermore, when the external noise is too strong, the above problems become more prominent. To address these problems, this paper, based on the build-up characteristics of HIF current, proposes a novel method combining stochastic resonance (SR) using noise and variational mode decomposition (VMD). First, SR is used to detect the transient zero-sequence current in strong background noise and obtain the output current. Second, VMD is adopted to decompose the output current, and low-frequency and high-frequency intrinsic mode functions (IMFs) are obtained. Then, kurtosis and skewness values are calculated to construct the starting and identification criteria, respectively. Finally, when the kurtosis value is greater than 3.0, the protection device is started, and then, when the skewness value is less than 0, that is, when negative skewness occurs, it is judged as a HIF; conversely, when the skewness value is greater than 0, that is, when positive skewness occurs, it is judged as CS, LS or IC. A large number of tests show that the method can realize accurate identification in strong background noise with an SNR = −10 dB.

When a single-phase-to-ground (SPG) fault occurs in a distribution network, the current is weak, and the fault conditions are complex; the existing fault detection technologies do not show reliable performance. Unlike the existing detection methods, we introduced a fundamental component shift and multiple-transient-feature fusion method in this paper. To remove the fundamental component (FC) that contained in the transient zero-sequence current (TZSC), Hilbert transform was used to calculate the analytic signal of TZSC. Then, the shift factor was introduced to remove the FC such that all the transient features can be kept. Second, three typical transient features indexes were calculated. Then, the multiple evidence estimation method was used to fuse the transient features indexes. Finally, selected the feeder with the maximum fault degree as the faulty feeder. To verify the accuracy and extensive applicability of the proposed method, developed the sample radial distribution network model with different fault conditions, arc fault model, and the modified IEEE-34-node test system with power electronics generator and wind energy conversion systems, the proposed method shows good performance.