Fusion of Signal Processing Techniques to Design Current and Voltage Features Based Protection Scheme for Utility Grid With Renewable Energy Penetration

This paper used fusion of Hilbert transform (HT), Stockwell transform (ST) and Alienation coefficient to design a protection scheme for utility network with high concentration of renewable energy (RE) generation. A hybrid fault detection index (HFDI) is designed which detects the fault events utilizing the features extracted from both current and voltage signals. Peak magnitude of HFDI is compared with a threshold magnitude (HTM) to detect fault events and discriminate such events from operational events. Faults are classified based on faulty phase numbers and a hybrid ground fault index (HGFI). Zero sequence voltage and zero sequence current are processed by application of ST to compute HGFI which effectively detects presence of ground during fault event. Performance of protection scheme is validated using utility grid of IEEE-13 bus test system where 50% RE penetration is used. Further, performance is also tested to detect faults incident on a practical utility network with RE penetration of 57%. Performance of algorithm is evaluated to recognize different fault events and operational events. Protection scheme is effectively tested to detect faults with high noise levels of 20 dB SNR (signal to noise ratio). Proposed protection scheme perform better compared to Alienation coefficient based protection scheme reported in literature.


HH-index Hybrid
The micro-grids (MG) and distributed generators are recent movements in the energy sector. Use of fossil fuels is continuously decreasing and renewable sources are considered as an effective, optimal and efficient way for future energy demand. Wind turbines and solar photovoltaic (PV) systems are most prominent sources of energy which are interfaced to the utility grids in the form of distributed generators (DG) [1]. DGs inject real and reactive power into the network of utility grid. This helps to reduce loss, increase grid resilience, and mitigate voltage sag during period of fault events. Apart from these benefits, high RE integration levels causes challenges in terms of protection due to power reversal in feeders, harmonic injections by converters, and fast ramp up/down rate due to generation uncertainty [2]. Signal processing techniques like Hilbert Transform (HT), Kalman filter (KF), Stockwell transform (ST), Gabor-Wigner transform (GWT), Mathematical Morphology (MM), Wavelet Transform (WT), Time-Time transform (TTT), Alienation coefficient, and Empirical Mode Decomposition (EMD) can effectively be used to design fault protection schemes of utility grids with RE integration [3]. A fault detection technique derived by data fault location using Gaussian process regression applied to a smart AC micro-grid is designed in [4]. In [5], authors introduced a detailed study investigate the impacts of DG integration into the grids on voltage profile, fault detection schemes and fault current levels. In [6], authors designed differential scheme of fault detection and classification using Maximal Overlap Discrete Wavelet Transform (MODWT). This protection scheme effectively applied to the micro-grid with DG sources. In [7], authors presented a study on sensor fault detection and isolation scheme and applied the same to detect fault events incident on interconnected smart network of power systems with high share of RE and feeding power electric vehicle (EV) charging stations. In [8], authors applied fuzzy system to design a protection approach for detection and classification of the short circuit fault incident on transmission line used for integration of wind power plant (WPP) to network of utility grid. A Pearson Correlation Coefficient (PCC) and support vector machine (SVM) based algorithm to detect high impedance arcing faults incident on micro-grids interfaced with DGs is introduced in [9]. Protection scheme is fast and detect faults with high accuracy. In [10], authors designed a protection approach for low impedance faults incident on micro-grids using flow direction of active power, voltage sag magnitude and current magnitude. Approach is effective to provide protection against low impedance faults of all types incident on micro-grids of every topology, configuration and every mode of operation. Detailed review and analysis of above discussed protection techniques indicates that fusion of signal processing techniques to extract features of both current and voltage signals can be used for improvement in performance of the protection schemes with high RE concentration. This is identified as fault detection research gaps and considered in this study. Key research contributions of this paper are detailed below: • Fusion of Signal Processing Techniques such as HT, ST, and alienation coefficient is used to design a protection scheme using features of both current and voltage signals which is effective to detect fault events incident on network of Utility with RE concentration.
• A HFDI is designed which is effective for detection of different fault events with RE presence. This is performed by comparing peak magnitude of HFDI with a threshold value.
• A HGFI is formulated by processing the zero sequence voltage and zero sequence current applying the ST to detect presence of ground with a fault. Faults are classified considering faulty phase numbers and HGFI.
• Performance of designed protection scheme is effectively validated using utility grid of IEEE-13 bus test system where 50% RE penetration is used. Further, performance is also tested to detect faults incident on a practical distribution network with RE concentration of 57%.
• Algorithm works well for various case studies of fault events and discriminates faults from operational events. Protection scheme is also effective to detect faults with high noise levels of 20dB SNR. VOLUME 10, 2022 • Protection scheme is superior compared to Alienation coefficient based protection scheme reported in literature.
The structure of this paper includes nine sections. Review of existing literature, research gaps and contribution in the research field are detailed in Section I. Test utility network interfaced with RE generators and implemented for validation of study is elaborated in Section II. Designed protection algorithm is described and illustrated in Section III. Results of simulation to describe fault detection are included in Section IV. Validation of protection algorithm on different cases of study are elaborated in Section V. Discrimination of fault events from the operational events is described using simulation results in Section VI. Testing of protection scheme on the practical utility grid is detailed in Section VII. Performance comparative analysis is described in Section VIII. Research conclusions are included in Section IX.

II. TEST UTILITY GRID INTERFACED WITH RE GENERATORS
Test utility grid is realized by integration of a solar power plant (SPP) rated at 1MW capacity and a WPP rated at a capacity of 1.5 MW to IEEE-13 nodes test system as described in Fig. 1. Test utility grid is rated at 5MVA and operated 60Hz frequency, voltages of 0.48 kV & 4.16 kV . All nodes are rated at 4.16 kV except node 634 which is operated at 0.48 kV . Hence, RE penetration level of 50% is used in this study. Utility grid transformer (UGTRF) is used to integrate the test network to the large area utility grid which is operated at 115kV [11]. SPP rated at 1MW is interfaced on 680 node with the help of STRF transformer and modelled using the parameters reported in [12]. WPP rated at 1.5 MW is integrated on node 680 with the help of WTRF transformer. WPP is modelled using parameters reported in [13]. Node 671 of test utility grid is taken as dault incidence node. Current and voltage are monitored and recorded at node 650 of test grid which is designated as hybrid protection scheme location (HPSL) point. Distribution transformer (DTRF) is used to operate node 634 at 0.48 kV and all other nodes are operated at 4.16 kV . Transformer details are provided in Table 1. Load data, capacitor data and feeder data available in [14] are utilized in this study.

III. PROPOSED HYBRID PROTECTION SCHEME FOR UTILITY GRID WITH RE CONCENTRATION
Proposed algorithm for fault detection and classification using fusion of signal processing methods to design a protection scheme for utility grid with RE concentration is enumerated in this section.

A. DETECTION OF FAULT EVENTS
A hybrid fault detection index (HFDI) is designed to detect the fault events which is illustrated in Fig.2. It is based on the processing of both current and voltage waveform using ST, HT and alienation coefficient. HFDI, hybrid Hilbert index (HH-index), hybrid Stockwell index (HS-index) and alienation index (A-index) are detailed in this section.

1) HYBRID HILBERT INDEX
The hybrid Hilbert index (HH-index) is formulated by processing both current and voltage signals by application of HT with sampling frequency of 3.84 kHz. Current (i(t)) is processed by application of HT for computation of absolute values of output matrix which is designated as HT current index (HCI) [15].
Here, PV , t, and τ indicate the Cauchy's principle value integral, time, and time period respectively. Voltage (v(t)) is processed using HT as detailed below to compute absolute values of output matrix which is considered as HT voltage index (HVI) [15].
HH-index is computed by multiplication of HCI and HVI as detailed below:- High magnitude during fault period and least effect of noise are main advantages of HH-index.

2) HYBRID STOCKWELL INDEX
Process the current signal (i(t)) using ST considering 3.84 kHz sampling frequency to compute output matrix with absolute values (STAMI) as detailed below [16], [17].
Here, t, f , and g(t) indicate spectral localization time, Fourier frequency, and Gaussian window function respectively.  Compute summation of every element of a column of the matrix STAMI and then take co-variance for computing co-variance factor (COVF) as described below.
Compute summation of all elements in a column of the matrix STAMI with the application below detailed relation.
Process the voltage signal (v(t)) using ST with considering 3.84 kHz sampling frequency and output matrix with absolute values (STAMV) is computed as detailed below [16], [17].
Compute summation of all elements in a column of the matrix STAMV applying the below detailed relation.
Compute median of the matrix STAMV applying below detailed relation.
Hybrid Stockwell index (HS-index) is computed by multiplication of VAI, VAV, MV and COVF using following relation.

3) ALIENATION INDEX
Following relation is applied to compute correlation coefficient (CORC) of current samples i 1 and i 2 at an interval of quarter cycle.
Here, Ns is the total samples in a cycle (Ns=64 are considered in this study); i 1 : current sample at time t 0 , i 2 current sample at time −T + t 0 , T: time period of current waveform. The alienation index (A-index) is derived from CORC applying the below detailed relation.

4) HYBRID FAULT DETECTION INDEX
HFDI is evaluated by multiplication of HH-index, HS-index, A-index and a hybrid weight factor (HWF).
Hybrid threshold magnitude (HTM) of 20000 is considered for HFDI to detect the fault events. This also discriminate the faults from healthy events. During faulty events, the magnitude HFDI is higher compared to the HTM and during the operational events this magnitude is lower compared to HTM.

B. FAULT EVENT CLASSIFICATION
Fault classification algorithm is described in Fig. 3. Fault events are classified and discriminated from each other using faulty phase numbers. To categorize the two phases fault and two phases to ground fault, a hybrid ground fault index (HGFI) is designed. HGFI also discriminates the three phase fault event with and without ground involvement. HGFI is computed by processing both the zero sequence voltage and current applying the ST. Following procedures is used to compute HGFI.
Calculate zero sequence current (I 0 ) using currents of all phases.
Calculate zero sequence voltage (V 0 ) using voltages of all phases recorded on node 650 of test grid.
Decompose zero sequence current using ST and obtain output matrix as detailed below.
Decompose zero sequence voltage using ST and obtain output matrix as detailed below.
Compute co-variance of summation of every column of STZI to compute zero sequence current co-variance factor (COVFZI).
Sum all the elements of a column of STZV matrix (SZI) as detailed below.
Sum all the elements of a column of STZV matrix (SZV) as detailed below.
Compute hybrid ground fault index by multiplication of MZI, MZV, SZI, SZV, COVFZI and weight factor (ZWF) as detailed below:- This study considered the Zero sequence weight factor (ZWF) qual to 10 6 . A hybrid zero sequence threshold (HZT) equal to 50 is considered to investigate the ground involvement during the fault. Magnitude of HGFI higher than HZT identifies the ground involvement during fault. Magnitude of HGFI lower than HZT indicates that ground is not involved during fault event.

IV. DETECTION OF FAULTS: SIMULATION RESULTS
This section discussed the investigated fault events including phase-A to ground fault (AGF), phase-A and phase-B fault (ABF), phase-A and phase-B fault with involvement of ground (ABGF), three phase fault (ABCF), and three phase fault with ground (ABCGF).

A. AGF EVENT
An AGF event is incident on 671 node of test network at 6 th cycle. Current and voltage are measured on HPSL point of test system and detailed in Fig. 4 (a) and (b) respectively. These current and voltage are processed applying HT for computation of HH-index which is elaborated in Fig. 4 (c). Current and voltage are also processed using ST to computed HS-index which is shown in Fig. 4 (d). Current is processed using Alienation coefficient for computation of A-index which is detailed in Fig. 4 (e). Proposed HFDI is computed by multiplying the HH-index, HS-index, A-index and HWF which is illustrated in Fig. 4 (f). Time of fault detection and peak magnitude of HFDI for AGF event are included in Table 2. Fig. 4 (a) indicates that current of fault phase-A is increased due to fault incidence at 6 th cycle. Fig. 4 (b) indicates that voltage of all phases decreased by small magnitude due to fault occurrence. However, decrease in magnitude of voltage of faulty Phase-A is more. Fig. 4 (c) indicates that HH-index magnitude of faulty phase-A increases due to fault occurrence at 6 th cycle. Further, magnitude of HH-index associated with healthy phases-A & B is increased slightly. Fig. 4 (d) indicates that HS-index magnitude of all phases has increased due to fault occurrence at 6 th cycle. Further, magnitude of faulty phase-A is more pronounced compared to healthy phases. Fig. 4 (e) indicates that A-index magnitude of all phases has increased to unity due to fault occurrence at 6 th cycle. Fig. 4 (f) indicates that HFDI magnitude of faulty phase is increased and becomes high in comparison to threshold. However, magnitude of healthy phases-B & C is low in comparison to threshold. Hence, AGF event has been effectively identified using the proposed algorithm. Table 2, indicates that AGF event is detected in small time interval of 1.1 × 10 5 s.

B. ABF EVENT
An ABF event between phases-A & B is incident on 671 node of test system at 6 th cycle. Current and voltage are measured on 650 node of test system and depicted in Fig. 5 (a) and (b) respectively. These current and voltage are processed using HT for computation of HH-index which is depicted in Fig. 5 (c). Current and voltage are also processed using ST for computation of HS-index which is shown in Fig. 5 (d). Current signal is processed using Alienation coefficient and A-index is computed which is detailed in Fig. 5 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is illustrated in Fig. 5 (f). Time of fault detection and peak magnitude of HFDI of ABF event are included in Table 2. Fig. 5 (a) indicates that current of fault phases-A & B is increased due to fault incidence at 6 th cycle. Fig. 5 (b) indicates that voltage of all phases decreased by small magnitude due to fault occurrence. However, decrease in magnitude of voltage of faulty phases-A & B is more. Fig. 5 (c) indicates that HH-index magnitude of fault phases-A & B has pronounced due to fault occurrence at 6 th cycle. Further, magnitude of HH-index corresponding to healthy phase-C is increased slightly. Fig. 5 (d) indicates that HS-index magnitude of fault phases-A & B has pronounced due to fault occurrence at 6 th cycle. However, magnitude of healthy phase-C increased by small amount compared to faulty phases. Fig. 5 (e) indicates that A-index magnitude of all phases has increased due to fault occurrence at 6 th cycle. Fig. 5 (f) indicates that HFDI magnitude of faulty phases-A & B is increased and becomes higher compared to threshold. However, magnitude of healthy phase-C is lower relative to threshold. Hence, AGF event is effectively identified using the proposed algorithm. Table 2, indicates that AGF event is detected in small time interval of 5 × 10 6 s.

C. ABGF EVENT
An ABGF event between phase-A and phase-B with ground involvement is incident on 671 node of test network at 6 th cycle. Current and voltage are measured on 650 node of test network and illustrated in Fig. 6 (a) and (b) respectively. These current and voltage are processed applying HT for computation of HH-index which is depicted in Fig. 6 (c).   Current and voltage are also processed using ST to computed HS-index which is shown in Fig. 6 (d). Current is processed using Alienation coefficient for computation of A-index which is detailed in Fig. 6 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is illustrated in Fig. 6 (f). Time of fault detection and peak magnitude of HFDI of ABGF event are included in Table 2. Fig. 6 (a) indicates that current of faulty phases-A & B is increased due to ABGF event incidence at 6 th cycle. Fig. 6 (b) indicates that voltage of all phases decreased by small magnitude due to ABGF event occurrence. However, decrease in magnitude of voltage of faulty phases-A & B is more. Fig. 6 (c) indicates that HH-index magnitude of faulty phases-A & B has increased due to ABGF event occurrence at 6 th cycle. Further, magnitude of HH-index corresponding to healthy phase-C is increased slightly. Fig. 6 (d) indicates that HS-index magnitude of faulty phases-A & B has increased due to ABGF event occurrence at 6 th cycle. However, increase in magnitude of healthy phase-C is small compared to faulty phases. Fig. 6 (e) indicates that A-index magnitude of all phases has increased due to ABGF event occurrence at 6 th cycle. Fig. 6 (f) indicates that HFDI magnitude of faulty phases-A & B is increased and becomes high relative to to threshold. However, magnitude of healthy phase-C is lower relative to threshold. Hence, ABGF event is effectively identified using the proposed algorithm. Table 2, indicates that AGF event is detected in small time interval of 1.0 × 10 6 s.

D. ABCF EVENT
An ABCF event between all phases is incident on 671 node of test system at 6 th cycle. Current and voltage are measured on 650 node of test system and depicted in Fig. 7 (a) and (b) respectively. These current and voltage signals are processed using HT for computation of HH-index which is elaborated in Fig. 7 (c). Current and voltage are also processed using ST for computation of HS-index which is shown in Fig. 7 (d). Current is processed using Alienation coefficient and A-index is computed which is detailed in Fig. 7 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is illustrated in Fig. 7 (f). Time of fault detection and peak magnitude of HFDI of ABCF event are included in Table 2. Fig. 7 (a) indicates that current of all phases is increased due to ABCF event incidence at 6 th cycle. Fig. 7 (b) indicates that voltage of all phases decreased by small magnitude due to ABCF event occurrence. Fig. 7 (c) indicates that HH-index magnitude of all phases has increased due to ABCF event occurrence at 6 th cycle. Fig. 7 (d) indicates that HS-index magnitude of all phases has increased due to ABCF event occurrence at 6 th cycle. Fig. 7 (e) indicates that A-index magnitude of all phases has increased due to ABCF event occurrence at 6 th cycle. Fig. 7 (f) indicates that HFDI magnitude of all phases is increased and becomes high relative to threshold. Hence, ABCF event is effectively identified using the proposed algorithm. Table 2, indicates that AGF event is detected in small time interval of 1.5 × 10 6 s.

E. ABCGF EVENT
An ABCGF event between all phases with involvement of ground is incident on 671 node of test system at 6 th cycle. Current and voltage are measured on 650 node of test system and depicted in Fig. 8 (a) and (b) respectively. These current and voltage are processed using HT for computation of HH-index which is depicted in Fig. 8 (c). Current and voltage are also processed applying ST for computation of HS-index which is shown in Fig. 8 (d). Current is processed using Alienation coefficient and A-index is computed which is detailed in Fig. 8 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is illustrated in Fig. 8 (f). Time of fault detection and peak magnitude of HFDI of ABCGF event are included in Table 2. Fig. 8 (a) indicates that current of all phases is increased due to ABCGF event incidence at 6 th cycle. Fig. 8 (b) indicates that voltage of all phases decreased by small magnitude due to ABCGF event occurrence. Fig. 8 (c) indicates that HH-index magnitude of all phases has increased due to ABCGF event occurrence at 6 th cycle. Fig. 8 (d) indicates that HS-index magnitude of all phases has increased due to ABCGF event occurrence at 6 th cycle. Fig. 8 (e) indicates that A-index magnitude of all phases has increased due to ABCGF event occurrence at 6 th cycle. Fig. 8 (f) indicates that HFDI magnitude of all phases is increased and becomes high relative to threshold. Hence, ABCGF event has been effectively identified using the proposed algorithm. Table 2, indicates that AGF event is detected in small time interval of 1.5 × 10 6 s.

F. FAULT CLASSIFICATION
Faults are categorized using faulty phase numbers. One phase and ground fault (AG, BG, CG) event is identified when HFDI is greater than threshold for one phase only. Two phase fault event with and without ground involvement (AB, BC, AC, ABG, BCG, ACG) is identified when HFDI is high relative to threshold for two phases. Ground involvement during two phase fault event is identified using HGFI as illustrated in Fig. 9. Magnitude of HGFI greater than hybrid zero sequence threshold (HZT) indicates that ground is fault part during the event of fault and ABG fault is discriminated from the AB fault event. Three phase fault event with and without ground involvement (ABC, ABCG) is identified when HFDI is greater than threshold for all three phases. Ground involvement during three phase fault is recognized using HGFI. Magnitude of HGFI will be greater than HZT due to ground involvement during three phase fault.

V. SIMULATION RESULTS: CASE STUDIES
Proposed protection algorithm is tested for different case studies to generalize the application and functionality for different scenario of grid.

A. FAULT INCIDENCE ANGLE VARIATIONS
An AGF event is incident on 671 node of test network at 6 th cycle for angle of fault incidence angles (AFI) of 0 • , 30 • , 45 • , 60 • , 90 • , 120 • , 150 • and 180 • to test functionality of protection algorithm for different angles of current and voltage waveform. Current and voltage are measured on HPSL point of test grid are processed applying protection algorithm to compute the HFDI. Peak magnitudes of HFDI corresponding to all investigated AFI is included in Table 3.
This has been observed from Table 3 that maximum value of HFDI for faulty phase-A has become high relative to threshold of 20000 corresponding to all investigated AFIs. The peak magnitude of HFDI for healthy phases-A & B is lower than threshold magnitude of 20000 corresponding to all investigated AFIs. Therefore, it is established that protection scheme functions effectively to detect fault events incident at different AFI on the voltage and current waveform.

B. FAULT IMPEDANCE VARIATIONS
An AGF event is simulated on node 671 of test utility grid at 6 th cycle taking fault impedance of 0.01 , 2 , 4 , 6 , 8 , and 10 to test functionality of protection algorithm for different fault impedance. Current and voltage are recorded on node 650 of test network are processed using protection algorithm to compute the HFDI. Peak magnitudes of HFDI corresponding to all investigated fault impedance is provided in Table 4. Table 4 indicates that maximum magnitude of HFDI for faulty phase-A has become high relative to threshold of 20000 corresponding to all investigated fault impedance. The peak magnitude of HFDI for healthy phases-A & B is lower than threshold of 20000 corresponding to all investigated fault impedance. Therefore, proposed protection algorithm functions effectively for detecting the fault events incident on voltage and current waveform with different fault impedance.

C. AGF EVENT INCIDENT AT DIFFERENT TEST NODES
An AGF event is incident on different nodes of test grid at 6 th cycle to test functionality of protection algorithm for different fault incidence nodes. Current and voltage are captured on 650 node of test grid are processed using protection algorithm to compute HFDI. Peak magnitudes of HFDI corresponding to all investigated fault incidence nodes is included in Table 5. Table 5 indicates that maximum magnitude of HFDI for faulty phase-A has become high relative to threshold of 20000 corresponding to all investigated fault incidence nodes. Peak magnitude of HFDI for healthy phases-A & B is lower than threshold of 20000 corresponding to all investigated fault incidence nodes. Therefore, it is established that protection algorithm functions effectively to detect fault events incident on different incidence nodes.

D. EFFECT OF NOISE
An AGF event is incident on 671 node of test grid at 6 th cycle. Current and voltage are recorded on 650 node of test grid. A noise of 20dB SNR is superimposed on both the current and voltage and illustrated in Fig. 10 (a) and (b) respectively. These noisy current and voltage are processed applying HT for computation of HH-index which is depicted in Fig. 10 (c). The noisy current and voltage are also processed applying ST to compute HS-index which is shown in Fig. 10 (d). Noisy current is processed using Alienation coefficient and A-index is computed which is detailed in Fig. 10 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is depicted in Fig. 10 (f). Fig. 10 (a) details that current of faulty phase-A is pronounced due to fault incident at 6 th cycle. Fig. 10 (b) elaborates that voltage of all phases decreased by small magnitude due to fault occurrence. However, decrease in magnitude of voltage of faulty Phase-A is more. Fig. 10 (c) indicates that HH-index magnitude of faulty phase-A has pronounced due to fault occurrence at 6 th cycle. Further, magnitude of HH-index associated with healthy phases-A & B is increased slightly. Fig. 10 (d) indicates that HS-index magnitude of all phases has increased due to fault occurrence at 6 th cycle. However, magnitude of faulty phase-A is more pronounced compared to healthy phases. Additionally, high magnitude spikes are also observed due to noise presence over entire time range. Fig. 10 (e) indicates that A-index magnitude of all phases has increased to unity due to fault occurrence at 6 th cycle. Small magnitude spikes are observed due to noise over complete time range. Fig. 10 (f) indicates that HFDI magnitude of faulty phase is increased and becomes high relative to threshold. Further, magnitude of healthy phases-B & C is lower relative to threshold. Hence, AGF event is effectively identified in noisy environment using the proposed algorithm.

VI. BEHAVIOR OF PROTECTION SCHEME WITH OPERATIONAL EVENTS
Simulation results for discriminating the faulty events from operational events are discussed in this section.

A. LOAD OPERATION
A load of 843kW and 462kVAr connected on 675 node of test network is switched off at 4 th cycle and reconnected to grid at 8 th cycle. Current and voltage are measured on 650 node of test network and illustrated in Fig. 11 (a) and (b) respectively. These current and voltage are processed applying the HT and HH-index is derived which is depicted in Fig. 11 (c). Current and voltage are also processed using ST to computed HS-index which is shown in Fig. 11 (d). Current is processed using Alienation coefficient and A-index is derived which is    Fig. 11 (e). Proposed HFDI is computed by multiplying the HH-index, HS-index, A-index and HWF which is illustrated in Fig. 11 (f). Fig. 11 (a) indicates that current of all phases has decreased between 4 th cycle to 8 th cycle. Fig. 11 (b) indicates that voltage of all phases remains unaffected. Fig. 11 (c) details that HH-index magnitude of all phases has decreased between 4 th cycle to 8 th cycle. Fig. 11 (d) indicates that HS-index magnitude of all phases has high magnitude peaks at 4 th cycle to 8 th cycle. Further, magnitude of peak at 8 th cycle is relatively high. Fig. 11 (e) indicates that A-index magnitude of all phases has peaks of unit magnitude at 4 th cycle and 8 th cycle. Fig. 11 (f) indicates that HFDI magnitude of all phases is lower relative to threshold. Hence, load operation event is considered as non-faulty event and differentiated from faulty events effectively using the proposed algorithm.

B. CAPACITOR OPERATION
A capacitor rated at 600kVAr connected on 675 node of test system is switched off at 4 th cycle and reconnected to grid at 8 th cycle. Current and voltage are measured on HPSL point of test grid and illustrated in Fig. 12 (a) and (b) respectively. These current and voltage are processed applying the HT and HH-index is derived which is depicted in Fig. 12 (c). The current and voltage are also processed using ST to computed HS-index which is shown in Fig. 12 (d). Current is processed using Alienation coefficient and A-index is derived which is detailed in Fig. 12 (e). Proposed HFDI is computed by multiplying the HH-index, HS-index, A-index and HWF which is illustrated in Fig. 12 (f). Fig. 12 (a) indicates that current of all phases has increased between 4 th cycle to 8 th cycle. A current spike is also incident at capacitor switching on instant. Fig. 12 (b) indicates that voltage magnitude of all phases remains same but low magnitude voltage spikes are observed due to switching on the capacitor at 8 th cycle. Fig. 12 (c) indicates that HH-index magnitude of all phases has increased between 4 th cycle to 8 th cycle. Fig. 12 (d) indicates that HS-index magnitude of all phases has high magnitude peaks at 4 th cycle to 8 th cycle. Further, magnitude of peaks at 8 th cycle is relatively high. Fig. 12 (e) indicates that A-index magnitude of all phases has peaks of high magnitude at 4 th cycle and 8 th cycle. Fig. 12 (f) indicates that HFDI of all phases is low relative to threshold at instant of switching on and off the capacitor. Hence, load operation event is considered as non-faulty event and differentiated from fault events effectively using the proposed algorithm. VOLUME 10, 2022

C. FEEDER OPERATION
Test grid feeder of 692 and 675 nodes is tripped at 4 th cycle and charged again at 8 th cycle. Current and voltage are measured on 650 node of test grid and depicted in Fig. 13 (a) and (b) respectively. These current and voltage are processed using HT and HH-index is derived which is depicted in Fig. 13 (c). Current and voltage are also processed using ST to computed HS-index which is shown in Fig. 13 (d). Current is processed using Alienation coefficient and A-index is derived which is detailed in Fig. 13 (e). Proposed HFDI is computed by multiplication of HH-index, HS-index, A-index and HWF which is depicted in Fig. 13 (f). Fig. 13 (a) indicates that current of all phases has remains un-affected between 4 th cycle to 8 th cycle. High magnitude current spikes are observed at moment of feeder re-closing. Fig. 13 (b) indicates that voltage magnitude of all phases remains same but low magnitude voltage spikes are observed due to switching on the feeder at 8 th cycle. Fig. 13 (c) indicates that HH-index magnitude of all phases remains same. However, high magnitude oscillations are observed due to feeder re-closing after 8 th cycle. Fig. 13 (d) indicates that HS-index magnitude of all phases has high magnitude peaks at 4 th cycle and 8 th cycle. Further, magnitude of peaks at 8 th cycle is relatively high. Fig. 13 (e) indicates that A-index magnitude of all phases has peaks of high magnitude at 4 th cycle and 8 th cycle. Fig. 13 (f) indicates that HFDI of all phases is lower relative to threshold at feeder tripping and re-closing instants. Hence, load operation event is considered as non-faulty event and differentiated from faulty events effectively using the proposed algorithm.

VII. TESTING OF PROTECTION SCHEME ON PRACTICAL UTILITY GRID
Distribution network of a practical utility grid rated at 11 kV is considered to validate performance of proposed protection schemes in real time scenario which is detailed in Fig. 14. This distribution network is fed from the 33/11 kV sub-station (SS) of sub-transmission system. Sub-station transformer is rated at 13 MVA, 33/11 kV. Bus B-1 is operated at 11 kV from where three feeders (Feeder-A, Feeder-B & Feeder-C) are emanating which feed the load in the a Phalodi region of India. Loads fed by distribution transformers (DT) are represented as concentrated load on the DT. All DTs are rated as 11/0.44kV. Five distribution transformers represented as ADT-1 to ADT-5 are connected to Feeder-A. Loads connected to these DTs are represented as AL-1 to AL-5 respectively. Three distribution transformers represented as BDT-1 to BDT-3 are connected to Feeder-B. Loads connected to these DTs are represented as BL-1 to BL-3 respectively. Four distribution transformers represented as CDT-1 to CDT-4 are connected to Feeder-C. Loads connected to these DTs are represented as CL-1 to CL-4 respectively. Details of loads and DTs is included in Table 6. All feeders use the aluminum conductor steel reinforced (ACSR) Weasal conductor rated at 11 kV having current carrying capacity of 114A [18]. Four solar power plants (SPP) designated as SPP-1 to SPP-2 are connected to the feeders using solar transformers (ST) ST-1 to ST-4 respectively. One wind power plant (WPP) is also connected to node B-13 using Wind transformer (WT-1). Details of SPPs and WPP with their transformers are included in Table 7. Incomer of 33/11kV transformer on the bus B-1 is considered as the HPSL for data measurement.
Fault data are recorded from the disturbance recorder installed at the 33/11kV Sub-station. Data of voltage and current corresponding to an AGF event incident on the node B-5 of test feeder-A are processed applying the proposed protection algorithm and HFDI is derived which is illustrated in Fig. 15. Peak HFDI associated to phases-A, B & C are computed as 3.3868 × 10 6 , 8910.8, and 11264 respectively. It is inferred that peak of HFDI for faulty phase-A is high relative to threshold HTM and it is low relative to healthy phases-B & C. Hence, AGF event observed on a real time distribution network is detected effectively.

VIII. PERFORMANCE COMPARATIVE ANALYSIS
To generalize the proposed protection approach, performance of algorithm is compared with alienation based protection algorithm reported in [19]. Comparative analysis of alienation based protection schemes and proposed approach considering different case studies and parameters is described Table 8. Investigated cases are indicated by Examined and cases which are not considered are indicated as Not examined. It is concluded that proposed protection schemes is more generalized compared to the algorithm introduced in [19]. Further, performance of alienation based method is affected for high noise levels and effective upto noise level of 40dB SNR. However, algorithm considered in this paper works well for noise levels of 20dB SNR. Further, proposed approach is more generalized for utility grids operating at high renewable energy share.

IX. CONCLUSION
A protection schemes for utility grid with high penetration of RE generation based on feature computed from voltage and current signals using HT, ST, and Alienation coefficient to detect fault events with the help of HFDI is presented in this paper. Fault classification is achieved using faulty phase numbers and a HGFI. It is concluded that protection schemes effectively detect fault events such as AGF, ABF, ABGF, ABCF, and ABCGF incident on a utility grid of IEEE-13 bus test system with 50% RE penetration. Algorithm is effective to detect faults incident at different nodes of test grid, different fault impedance, and different angles of fault incidence. Protection scheme is effective to detect faults with high noise levels of 20dB SNR. Protection scheme is effective to discriminate the operational events from the faulty ones. This is also concluded that protection scheme effectively detects the faults incident on a practical distribution network with RE penetration of 57%. This is also concluded that protection scheme is better relative to Alienation coefficient based protection scheme reported in literature.