Table 5 Distinctive features of investigated signal processing-based protection schemes
References | Year | Citation | DER | Required measures | Method | Features |
|---|---|---|---|---|---|---|
[110] | 2016 | 20 | Synchronous and inverter based | Fault current | WT and decision tree | Twelve statistical features such as: mean, standard deviation, energy, entropy, etc., are extracted from fault current decomposition to train the decision tree High impedance faults are detectable Requires offline training, and high computational burden |
[111] | 2016 | 260 | Inverter based | Actual and sequence components of fault current | WT and decision tree | Nine features are used for fault detection, while fifteen for fault classification 70% of input data are used for training, remaining 30% for testing High computational burden due to required trainings Low-impedance faults are only used for training |
[112] | 2021 | 10 | Synchronous and inverter based | Voltage and current data | WT and random forest | Random forest is used as a data mining tool to accurately process a large input database, unlike the decision tree. 75% of input data are used for training, the remaining for testing Considers DERs outages and fault initiation periods Robust against measurement noise Requires high capability software for training |
[113] | 2017 | 35 | Synchronous and inverter based | Voltage or current data | WT and park’s transformation | The d-q voltages/currents feed wavelet model Not preferable for high-impedance faults High sampling frequency, and low accuracy Large time response for data processing Detection signal is delayed to distinguish false faults |
[115] | 2019 | 19 | Not reported | Voltage and current waves | TWs | Fault is detected based on traveling waves polarities Considers zero-sequence voltage to avoid false detection Considers fault inception time, type, and resistance Applicable for SLG faults in non-effectively grounded systems |
[116] | 2014 | 123 | Inverter based | Fault current wave | TWs | Low-bandwidth communication is employed for high-speed operation Stable during normal transients i.e. motor starting Considers traveling wave amplitude, timing, and polarity for accurate detection |
[117] | 2017 | 5 | Synchronous and inverter based | Local currents and fault current wave | TWs | Detect fault based on WT of the traveling wave, while zone classification relies on wave signs Applicable for close-in faults Stable during switching transients and external abnormalities |
[121] | 2022 | 2 | Synchronous and inverter based | Currents at both ends of line | S-transform-based differential current | Varied threshold value with the operating mode and fault impedance, i.e. high impedance fault. High impedance faults are detectable Robust against measurement noise |
[122] | 2021 | 10 | Inverter based | Current and voltage data | S-transform-based distance relay | Fault energy is used as a fault indicator, while distance relay defines trip timings. Low computational burden High impedance faults are addressed |
[123] | 2014 | 15 | Synchronous and inverter based | Currents at both ends of line | S-transform and decision tree | Low computational burden Fast response (1–1.5) cycle Requires offline training |
[126] | 2021 | 8 | Synchronous and inverter based | Currents at both ends of line | Hilbert–Huang transform | Low required time for fault detection and classification processes Limited to fault impedance larger than 1000 Ω Self-adaptive threshold: large in normal conditions and decreases with faults |
[82] | 2018 | 144 | Synchronous and inverter based | Current measurements | Hilbert–Huang transform | Three distinctive differential features are used: phase current energy, standard deviation of phase current, and zero-sequence current energy Applicable for high-impedance faults Machine learning model uses 70% of input data for training, remaining 30% for testing Offline training is needed |
[127] | 2008 | 120 | Inverter based | Voltage data | Harmonic content-based | THD value is dependent on network configuration Individual values of THD are used to classify fault type Applicable only for identical DERs High impedance faults are not investigated |
[128] | 2016 | 13 | Inverter based | Harmonic current (5th harmonic) | Harmonic content-based | Inverter-based DER injects harmonic currents Not applicable for high-impedance faults Inaccurate in a harmonic rich system |
[129] | 2022 | 1 | Inverter based | Multiple Harmonic components | Harmonic content-based | Multiple harmonics injection ameliorates sensitivity Reliable and low-cost due to communication-free protocol Injected harmonic component has a magnitude of 10% of fault current Each inverter injects a distinct harmonic content Detects high-impedance faults |
[130] | 2018 | 4 | Inverter based | Voltage and current data Harmonic current | Harmonic content-based | Optimized coordination settings using Particle Swarm Optimization Only low-impedance faults are verified Avoids resonance conditions when selecting the injected harmonic |