Identification of Intermittent Ground Fault Lines in Medium Voltage Shipboard Power Systems

To locate increasingly frequent intermittent ground faults in medium-voltage shipboard power systems, this study proposes a fault branch selection method using wavelet analysis. By analyzing transient fault current characteristics in ship medium-voltage grids, this method utilizes wavelet analysis tools to extract features and obtain modulus maxima of branch currents. Based on a dual threshold detection strategy, the fault branch is then identified. The feasibility and validity of the proposed method are verified through simulations on the Matlab/Simulink platform. The simulation results demonstrate the capability of the wavelet analysis approach to accurately locate intermittent ground faults in ship medium-voltage power systems. This research provides an effective solution for fault branch selection of intermittent grounding faults in ship medium-voltage grids.


Introduction
As shipboard power systems transition from low to medium voltage levels, intermittent insulation faults absent or rare in low-voltage AC networks become increasingly prevalent under medium voltage conditions [1].Major intermittent faults in ship grids include intermittent arc grounding faults and virtual cable faults resulting from chassis vibrations.However, existing insulation monitoring devices poorly diagnose such faults and are prone to false alarms.Moreover, intermittent ground faults often precede the deterioration of network insulation.Without timely monitoring of intermittent faults, insulation degradation may progress to metallic grounding faults [2].
To address the impacts of intermittent ground faults in ship medium voltage grids, this study analyzes transient current characteristics.It proposes a dual threshold detection method using wavelet analysis for fault branch identification.The proposed method is validated through simulations in Matlab/Simulink.Results confirm the capability of the wavelet-based technique to accurately locate intermittent ground fault branches in ship medium voltage power systems.This research provides an effective solution for fault branch identification of intermittent shipboard grounding faults.

Analysis of transient current characteristics
When an insulation fault occurs in the medium voltage power grid of a ship, the phase voltage of the faulty phase will suddenly drop, resulting in transient discharge of the system's distributed capacitance to the ground and forming a transient discharge current.Conversely, at the moment of insulation failure, the phase voltage of the non-fault phases undergoes a sudden rise, causing transient charging of the distributed capacitance to the ground and forming a transient charging current.The transient discharge and charging currents together constitute the overall transient current of the system during insulation faults [3][4][5].
The transient equivalent circuit diagram of insulation faults in the medium voltage neutral point high resistance grounding system of ships is depicted in Figure 1.In the equivalent circuit, C represents the sum of the phase-to-ground capacitance of each line.Parameters 0 L and 0 R represent the equivalent inductance and resistance, respectively, of the zero sequence current flowing through the circuit after insulation failure.Parameter 0 u represents the zero sequence voltage generated in the system after insulation failure.Parameter N R represents the neutral grounding resistance of the ship's medium voltage power system, which satisfies the following relationship [6]: m U  is the voltage amplitude of the phase voltage.
According to Kirchhoff's voltage law, the following equivalent relationship can be obtained , the fault branch will begin to experience a transient surge current, with a certain randomness in the characteristics of the current.After one or two power frequency cycles, the intense transient surge will disappear, forming a stable fault current.When 0 , the attenuation characteristics of the fault branch show a certain periodicity from the occurrence of large impulse current to the formation of steady-state fault current, and the attenuation time constant is Due to the unique nature of the ship power network, the distributed capacitance to the ground formed in the cables is much larger than that in the land power grid.This results in more intense transient characteristics of the system when insulation faults occur, with shorter duration, instantaneous, and strong impact.
The transient zero sequence current meets the following formula [7]: in the formula,

Cm
A is the amplitude of capacitive current; f  is the transient fault oscillation frequency;  represents the oscillation attenuation coefficient in the fault state, which satisfies the following relationship: When a single-phase insulation fault occurs, the expression for transient current is: In summary, the key transient current characteristics for single-phase insulation faults are [8]: (1) The transient zero sequence current has a much larger magnitude compared to the steady-state component.Relative to onshore grids, the shipboard transients are more intense, shorter in duration, and exhibit instantaneous sharp peaks.
(2) The transient and steady-state currents have consistent directionality.Non-fault branches show current flow from the bus to the feeder, while fault branches draw current from the feeder to the bus.
(3) The transient fault current magnitude depends on the fault inception angle.Larger fault closing angles produce greater transient amplitudes.
In essence, shipboard transients are high magnitude, rapidly damped, and unidirectional, with severity determined by the point-on-wave of fault occurrence.Analyzing these unique current signatures facilitates the development of detection algorithms optimized for maritime medium-voltage power systems.The transient behavior reflects the specific parameters and topology of the shipboard distribution network.

Double Threshold Detection Method Based on Wavelet Analysis
Intermittent ground faults exhibit transient and impulse characteristics, containing abundant highfrequency fault information.Therefore, transient signal-based line selection techniques can identify intermittent fault branches.
Wavelet analysis decomposes the transient current at fault occurrence, enabling detailed examination of high-frequency components.Wavelet packet decomposition further analyzes the signal to extract accurate transient information.Due to this capability, wavelet-based methods are widely adopted for fault location.For a faulty feeder, wavelet analysis of the zero sequence current yields the maximum modulus value with opposite polarity compared to healthy branches.Intermittent faults reproduce this distinct peak transient.
The proposed dual threshold detection exploits this behavior by comparing the wavelet modulus maximum against a peak threshold within a detection window.The number of thresholds exceeding events N is counted and compared to limit N0.If N exceeds N0, an intermittent fault is declared on that branch.
In summary, by leveraging unique transient signatures, the wavelet-based line selection algorithm reliably identifies intermittent fault branches.The dual threshold detection formalizes the transient analysis using modulus peak screening and occurrence counting, providing an effective approach for shipboard intermittent ground fault monitoring.
The main line selection process of the dual threshold detection strategy is shown in the figure:

Simulation verification
A ship medium voltage power system is modeled in Matlab/Simulink, as shown in Figure 3. Threephase generation and distributed feeder capacitance representing 6.6 μF balanced capacitance are simulated.Rn represents the system-neutral grounding resistor.Intermittent faults are introduced by controlling a transition resistance switch at various frequencies.
The three identical branches model healthy feeders.The fourth branch emulates a faulty case.Thus, by comparing the normal and faulty cases, the fault location algorithm can be validated for identifying the faulty branch based on unique signatures.This representative system establishes a simulation testbed to assess intermittent ground fault monitoring techniques optimized for shipboard medium voltage distribution networks.
By providing details on the model topology, parameters, and fault emulation approach, the simulation platform is established within the context of real shipboard power system characteristics.The description focuses on the key components and configurations necessary to evaluate the proposed fault detection method.The modular approach isolates the effects of intermittent faults to rigorously validate the line selection algorithm performance.

Simulation analysis of distributed capacitance balance in power grid
With balanced feeder capacitance and a 500 Ω transition resistance R, wavelet analysis is performed on the branch leakage currents for an intermittent fault event.Figures 4-5 show the resulting wavelettransformed current signatures under test case 1.
The wavelet decomposition clearly distinguishes the fault current transient in branch 4 from the healthy branches 1-3.The distinct polarity reversal and large modulus maximum confirm branch 4 as the faulty feeder.This demonstrates the capability of the wavelet technique to reliably discriminate the fault location based on the transient current signature when an intermittent ground fault occurs.Additional test cases further validate the robustness of the proposed approach [9][10].The consistent performance for both low and high-impedance intermittent faults demonstrates the robustness of the wavelet-based line selection approach.By exploiting the unique fault current transient profile, the algorithm accurately identifies the faulty feeder across a range of fault types common in shipboard power systems.

4.2
Simulation analysis of distributed capacitance imbalance in power grid To evaluate the effects of unbalanced capacitance, test case 3 models a 0.5 capacitive imbalance where phase B is reduced to 1.5 μF while phases A and C remain at 3 μF.With a 500 Ω transition resistance, wavelet analysis is performed on the branch leakage currents.This demonstrates the capability of the proposed technique to select the correct faulty branch under representative shipboard capacitive imbalance scenarios.By exploiting the fault transient current signature, the wavelet algorithm reliably locates insulation faults even with capacitive distribution in the medium voltage network.
Under test case 4 with a 20 kΩ transition resistance, wavelet analysis was performed on the branch leakage currents under 0.5 capacitive imbalance, as shown in Figures 10-11   Compared to the 500 Ω and 20 kΩ balanced cases, significant differentiation remains between the healthy and faulty branches despite the capacitive unbalance and high fault impedance.The large modulus maximum identifies the faulted feeder, validating the dual threshold detection methodology.With appropriate threshold selection, intermittent faults can be reliably located.
The results demonstrate the capability of the proposed wavelet-based approach to accurately select fault branches, even with grid capacitive imbalance.However, high transition resistance faults produce very weak zero sequence currents.This places demanding requirements on signal sampling to capture these elusive transients.
In summary, the technique robustly identifies fault location under representative shipboard capacitive imbalance and line impedance conditions.The need for sensitive sampling equipment is a practical implementation challenge for high-impedance intermittent faults.Further testing on a wide range of fault scenarios is recommended.

Conclusions and Prospects
With intermittent ground faults becoming increasingly prevalent in shipboard power systems, this paper puts forward a wavelet-based technique for faulty branch selection.The proposed method leverages wavelet analysis to extract transient current signatures and identify the maximum modulus values indicating fault location.A dual threshold detection strategy then screens the transient peaks to locate the faulty feeder.The approach is validated through Simulink simulations of balanced and unbalanced capacitance conditions.Results demonstrate accurate intermittent fault diagnosis across representative ship grid scenarios.This research presents an effective wavelet-domain line selection algorithm for locating intermittent ground faults in maritime medium voltage distribution networks.By exploiting the unique current transients during fault events, the method reliably discriminates the faulted branch from healthy feeders.The work addresses a pressing reliability challenge and provides a new solution for shipboard intermittent ground fault monitoring.

Figure 1 .
Figure 1.Transient Equivalent Circuit Diagram of Insulation Fault.

BeginFirstlyFigure 2 .
Figure 2. Flow Chart of Double Threshold Detection Strategy Line Selection.

Figure 4 .
Figure 4. Wavelet Analysis Waveform of Normal Branch under Test Case 1.

Figure 5 .
Figure 5. Wavelet Analysis Waveform of Fault Branch under Test Case 1.With a 20 kΩ transition resistance representing higher impedance faults, the wavelet transformed branch currents for test case 2 are depicted in Figures6-7.
Figure 5. Wavelet Analysis Waveform of Fault Branch under Test Case 1.With a 20 kΩ transition resistance representing higher impedance faults, the wavelet transformed branch currents for test case 2 are depicted in Figures6-7.

3 Figure 6 .
Figure 6.Wavelet Analysis Waveform of Normal Branch under Test Case 2.

Figure 7 .
Figure 7. Wavelet Analysis Waveform of Fault Branch under Test Case 2. Similar to the previous low impedance fault, the wavelet analysis clearly identifies the polarity reversal and large transient modulus peak corresponding to the faulty branch 4. Despite the higher resistance fault, the proposed technique reliably discriminates the fault current transient signature.

Figure 8 .
Figure 8. Wavelet Analysis Waveform of Normal Branch under Test Case 3.

Figure 9 .
Figure 9. Wavelet Analysis Waveform of Fault Branch under Test Case 3. As shown in Figures 8-9, despite the capacitive unbalance, the wavelet transform maintains clear identification of the fault branch 4 based on the transient polarity and large modulus.The performance is consistent with the previous balanced cases.This demonstrates the capability of the proposed technique to select the correct faulty branch under representative shipboard capacitive imbalance scenarios.By exploiting the fault transient current signature, the wavelet algorithm reliably locates insulation faults even with capacitive distribution in the medium voltage network.Under test case 4 with a 20 kΩ transition resistance, wavelet analysis was performed on the branch leakage currents under 0.5 capacitive imbalance, as shown in Figure 9. Wavelet Analysis Waveform of Fault Branch under Test Case 3. As shown in Figures 8-9, despite the capacitive unbalance, the wavelet transform maintains clear identification of the fault branch 4 based on the transient polarity and large modulus.The performance is consistent with the previous balanced cases.This demonstrates the capability of the proposed technique to select the correct faulty branch under representative shipboard capacitive imbalance scenarios.By exploiting the fault transient current signature, the wavelet algorithm reliably locates insulation faults even with capacitive distribution in the medium voltage network.Under test case 4 with a 20 kΩ transition resistance, wavelet analysis was performed on the branch leakage currents under 0.5 capacitive imbalance, as shown in

3 Figure 10 .
Figure 10.Wavelet Analysis Waveform of Normal Branch under Test Case 4.

Figure 11 .
Figure 11.Wavelet Analysis Waveform of Fault Branch under Test Case 4.Compared to the 500 Ω and 20 kΩ balanced cases, significant differentiation remains between the healthy and faulty branches despite the capacitive unbalance and high fault impedance.The large modulus maximum identifies the faulted feeder, validating the dual threshold detection methodology.With appropriate threshold selection, intermittent faults can be reliably located.The results demonstrate the capability of the proposed wavelet-based approach to accurately select fault branches, even with grid capacitive imbalance.However, high transition resistance faults produce very weak zero sequence currents.This places demanding requirements on signal sampling to capture these elusive transients.In summary, the technique robustly identifies fault location under representative shipboard capacitive imbalance and line impedance conditions.The need for sensitive sampling equipment is a practical implementation challenge for high-impedance intermittent faults.Further testing on a wide range of fault scenarios is recommended.