A Method for Fault Section Identification of Distribution Networks Based on Validation of Fault Indicators Using Artificial Neural Network
1
Digital Solution Laboratory, KEPCO Research Institute, Daejeon 34056, Republic of Korea
2
Department of Electrical Engineering, Kyungnam University, Changwon 51767, Republic of Korea
*
Author to whom correspondence should be addressed.
Energies 2023, 16(14), 5397; https://doi.org/10.3390/en16145397
Submission received: 22 June 2023
/
Revised: 10 July 2023
/
Accepted: 14 July 2023
/
Published: 15 July 2023
(This article belongs to the Special Issue Artificial Intelligence for Control Applications in Power and Energy Systems)
Abstract
:A fault section in Korean distribution networks is generally determined as a section between a switch with a fault indicator (FI) and a switch without an FI. However, the existing method cannot be applied to distribution networks with distributed generations (DGs) due to false FIs that are generated by fault currents flowing from the load side of a fault location. To identify the false FIs and make the existing method applicable, this paper proposes a method to determine the fault section by utilizing an artificial neural network (ANN) model for validating FIs, which is difficult to determine using mathematical equations. The proposed ANN model is built by training the relationship between the measured A, B, C, and N phase fault currents acquired by numerous simulations on a sample distribution system, and guarantees 100% FI validations for the test data. The proposed method can accurately distinguish genuine and false Fis by utilizing the ability of the ANN model, thereby enabling the conventional FI-based method to be applied to DG-connected distribution networks without any changes to the equipment and communication infrastructure. To verify the performance of the proposed method, various case studies considering real fault conditions are conducted under a Korean distribution network using MATLAB.
1. Introduction
As electricity demand increases and the power system becomes more complex, reliability in the distribution network has become important. To increase reliability, a distribution automation system (DAS) has been developed to monitor and control protective devices and switches. The DAS performs functions such as fault restoration, network analysis, and load balancing to operate the distribution system efficiently [1,2,3].
To improve the reliability of the power supply, the fault restoration function is very important. A fault restoration function is carried out in the order of determining the fault section i and restoring service. The fault section is determined by using fault information that is measured on feeder remote terminal units (FRTUs) when a failure occurs in the distribution network [4,5]. The service restoration function isolates the fault section from the main source and transfers the healthy sections to neighboring feeders through tie lines to supply the power back [6].
The method of determining the fault section differs depending on whether the power distribution network is grounded or not. When a fault occurs in an ungrounded power distribution network, the fault current is much smaller than the fault current in a grounded power distribution network so that the power can be supplied without any interruption even in a short circuit fault. However, the fault can affect the power system if the fault is left unattended for a long time.
There are three general methods for determining the fault section: sequential switching, binary search, and a comparison of the phase angles of the zero-sequence current. The sequential switching method is a method used to determine the fault section by sequentially closing automatic switches from the circuit breaker. This method has the problem with regard to the fact when a large area of outage occurs, there is a high possibility of switching failure occurring due to the large number of switching operations. The binary search method divides the network into two equal parts and searches for the fault location by sequentially opening the automatic switch at the 1/2 point of the network with the circuit breaker closing. This method has the disadvantage of the operating procedure lacking the ability to be calculated in advance and the location of the operating switches not being sequential. The method using a comparison of the phase angle of the zero-sequence current determines a fault section by comparing phase differences between inter-phase voltages measured at terminal devices and zero-sequence currents. This method requires transformers that can measure a zero-sequence current at the terminal device and a fault indicator (FI) function.
In Korean distribution networks, a fault section is determined by utilizing FIs generated from FRTUs. The FI is generated when a fault current flowing through it exceeds a certain threshold value, and the fault section can be determined by judging between a switch with an FI and a switch without any FIs [7]. However, the existing fault section identification method can only be applied when the fault current in the distribution network is unidirectional. It is difficult to apply the conventional method to distribution networks with a bidirectional flow due to distributed energy resources (DERs) and wye-delta transformers causing the reverse fault current flow. If there are switches with a false FI that has been affected by a reverse fault current, incorrect fault section identifications can be performed.
As a large number of DERs are integrated into the distribution network, solutions have been proposed to address voltage rise issues and the malfunctioning of protective devices due to the bidirectional network [7]. In particular, a solution has been proposed to replace existing protective devices with directional protective devices that have directional information in order to solve the malfunctioning of protective devices. However, in Korea, it is difficult to implement this solution because it would require a significant amount of cost and time to replace all existing non-directional protective devices, which are already in use.
Various studies on the methods used to identify faulty sections in grounding systems have been conducted. Some works propose impedance-based fault section identification methods [8,9,10,11]. These methods involve constructing an equivalent impedance from the voltage and current data measured at a substation in order to identify faulty sections. However, these methods are applied with difficulty under multiple fault situations. In [12], a method based on traveling waves is proposed. It takes advantage of protective devices that can measure traveling waves in the event of a fault in the distribution network in order to identify faulty sections. An injection-based fault section identification method that requires additional equipment for measuring injection signals is also presented in [13,14]. The methods used in [1] and [12] utilize low-cost distribution phase measurement devices and intelligent electronic devices (IEDs) to identify faulty sections, respectively.
A review of the previous studies confirms that equipment that has to be newly installed should be utilized, resulting in high costs and extensive time being required. In addition, the methods using traveling waves are applied with difficulty to distribution networks with a large number of branches requiring complex traveling wave analysis. However, the proposed method in this paper can accurately determine fault sections by distinguishing genuine and false FIs with the help of an artificial neural network (ANN) model, thereby enabling the conventional FI-based method to be applied to DG-connected distribution networks without any changes to the equipment and communication infrastructure.
This paper proposes a method of identifying faulty sections using an ANN model to validate FIs, which enables the use of the conventional FI-based fault section identification method without any additional installation of equipment. The structure of this paper is as follows: Section 2 explains the conventional fault section identification method and its limitations. Section 3 describes the proposed ANN-based fault section identification method and the overall training process of the proposed model. Section 4 verifies the performance of the proposed method through various case studies considering real fault conditions. Finally, the conclusion is presented in Section 5.
2. Conventional Fault Section Identification Method
When a fault occurs in a power distribution network, an FI is triggered by the fault current. The information from the FI is then sent to the DAS server, which determines the faulty section. The FI is triggered when a fault current greater than the predetermined threshold flows and the fault condition is maintained for a certain amount of time. The fault section is determined as the section between a switch with an FI and a switch without any FIs [7]. Figure 1 shows the situation in which a fault occurs in Section 3. The FI is triggered due to a fault current larger than the threshold value flowing through the CB, SW1, and SW2. The fault section can be determined as Section 3, since it is the area between the switch at which the FI is last detected and the switch at which the FI was not generated.
The FI of the load-side terminal device is not triggered when a fault occurs at a fault point in a single-phase distribution line since fault currents flow in only one direction. However, in a three-phase distribution line, if an unbalanced fault such as a single line to ground fault occurs, the FIs of the load-side terminal device at the fault point can be triggered. In addition, if a wye-delta transformer is connected to the distribution line, the FI of the load-side terminal device at the fault location can also be triggered when the current flows from the load side to the main source side. Figure 2 shows the zero-sequence circuit when a fault occurs in a distribution line connected to a load, a DER, and a wye-delta transformer. The magnitude of the current from the load side to the power side at the fault point is determined by the fault impedance, the loads connected to the load side, the capacity of the DER, and the wye-delta transformer. As the amount of load connected to the load side increases, the impedance decreases, which causes a rise in the current flowing from the load side to the fault location. In addition, the current flowing from the load side to the fault location increases as the capacity of the wye-delta transformer increases. In other words, the FI of the terminal device installed on the load side can be unintentionally activated if there is a heavy load or the transformer has a large capacity.
Figure 3 depicts the vector diagrams of the power side and the load side of a fault location when a single line to ground fault in phase A occurs. When the load is heavy on the load side, the power side measures a large fault current on the phase A fault location, and the vector diagram of the power side appears, as shown in Figure 3. Since most of the load-side phase A current flows to the fault location, the zero sequence current flows in the opposite direction to the power side, as shown in Figure 3. As the amount of load increases, the impedance on the load side decreases. Therefore, the magnitude of the zero sequence current on the load side increases, and the FI of the switch installed on the load side is generated. The FI on the load side is activated on only the neutral phase (phase N) because the phase A current is very small. In addition, when the power-side current of the wye-delta transformer is three-phase balanced, there is no zero sequence current on the power side and zero sequence current on the load side. However, if there is an imbalance fault, such as a ground fault in the power side, zero sequence current will occur. The zero sequence current generated on the power side is induced on the load side, and a current close to infinity flows since the internal impedance is close to 0. The current flowing close to infinity on the load side is induced back to the power side, and a large current flows from the load side to the power side.
Figure 4 illustrates a case in which a fault occurs in a distribution line that is connected to a load, a DER, and a wye-delta transformer. As shown in Figure 4, the conventional fault section identification method incorrectly identifies the fault section. The fault current is larger than the threshold value of the FI for CB, SW1, SW2, and SW3, so the FIs for all switches are activated. The existing fault section determination method cannot be used because the last switch with an FI detected is SW3, causing the fault section to be determined as Section 4.
3. Proposed ANN Model for FI Validation
3.1. Necessity of ANNs
An ANN is a model of human neuron cells, and it is used to understand the relationship between input and output by training input/output data [15,16,17,18]. The advantage of ANNs is that they are advantageous in solving problems by learning relationships between the input and output that are only mathematically expressed with difficulty.
Figure 5 is an enlarged representation of the FI information and current data distribution of the switches when a fault occurs in a distribution network connected to a heavy load or a DER using MATLAB, which has been widely used for engineering purposes [19,20]. In the figure, ‘FI’ represents the fault current flowing from the power side to the fault point, and ‘False FI’ refers to the fault current flowing from the load side to the fault location or when any FI is not detected. To distinguish between the FIs and false FIs without using ANNs, it is necessary to use a formula that represents the boundary between FIs and false FIs, as shown in Figure 5. However, there is a certain range in which the fault currents generate both genuine and false FIs, and it is difficult to express the boundary line between FIs and false FIs using a formula when the value of MAX (A, B, C) is between 200 (A) and 400 (A). The ANN can perform non-linear regression by using a hidden layer and an activation function with non-linear characteristics. Therefore, it is advantageous to use ANNs to discriminate between FIs and false FIs.
3.2. Fault Section Identification Scheme
In this paper, we propose a method that can determine the fault section by using the validation of FIs based on an ANN without replacing any of the equipment installed in the distribution network. Figure 6 shows the DAS equipped with the ANN. When a fault occurs, the equipment installed in the distribution line provides alarm information to the front-end processor (FEP). After determining the fault and the faulted line, the fault section identification application requests measurement information from the FEP, and determines the fault section based on the discrimination results of the FIs using the ANN. The ANN proposed in this paper is a network that receives the current magnitude of the three phases as input and determines the presence or absence of FIs.
3.3. Learning Model and Configuration of the ANN
The model is created by using three kinds of data. The data types are divided into training data for learning the relationship between input and output, validation data for selecting a better learning model each time, and test data for testing the performance of the generated model after the learning is completed.
Figure 7 shows a flowchart for the process of creating a learning model. Firstly, the ANN structure is created, and then the learning is performed using the train data. Each time learning is performed, the error rate is calculated using the validation data for the trained model. If the error rate is higher than the previous error rate, the previous learning model is selected. Otherwise, the current learning model is selected. When the error rate remains the same for a certain length of time or after a certain number of learning iterations, the learning is terminated and the process is ended.
The input and output layers of the ANN are composed of phase fault currents and the corresponding FIs. The ANN consists of an input layer, three hidden layers, and an output layer, as indicated in Figure 8. The number of nodes in each layer is set to 4, 64, 32, 16, and 1, respectively. The Relu function is used as an activation function between the input layer and each hidden layer, and between each hidden layer. As the value of the output layer should be either 0 or 1, the activation function used between the last hidden layer and the output layer is the sigmoid function. The nodes in the input layer are ×1, ×2, ×3, representing phase A, B, and C currents, and ×4 representing the phase N current. The node in the output layer is y1, with a value between 0 and 1 when using the activation function. The output value is determined to be a valid FI if it is between 0.5 and 1, and a no/false FI if it is between 0 and 0.5.
3.4. Generation and Verification of Data for the ANN
By using MATLAB/Simulink, training data are generated by changing the fault resistance, the distance from a substation to the fault location, the wye-delta transformer capacity, and the load affecting the magnitude of the fault currents. It is worth mentioning that the data for the ANN model are acquired using computer simulations since they provide faster results and a lower cost compared to experimental and analytical investigation [21,22,23]. The results from computer simulations should be validated using proper experimental and/or analytical results to ensure the validity of the results and increase confidence in the methodology [24,25,26]. The learning model is used to validate FIs when a fault occurs. Therefore, the training data for generating the learning model should consist of FI validation information and fault current data. A sample 22.9 kV distribution network is used, as shown in Figure 9. It consists of a 22.9 kV equivalent source, three line sections divided by SW1 and SW2, a lumped load, a wye-delta transformer, and a DG. The parameters used in the MATLAB/Simulink model are provided in Appendix A. A fault in Section 2 is assumed to obtain fault currents data for SW1 and SW2, representing switches on the source and load side, respectively. The length of Section 1 is changed to adjust the distance to the fault location. A load, n DER, and a transformer are placed at the end of Section 3, and a fault occurs in Section 2. Table 1 shows the input range of the parameters used to extract the training and test data. The fault resistance is set to a maximum of 30 (ohms), and the range in the fault distance is considered to be a maximum of 30 (km), since a Korean distribution line length is usually within 30 (km)]. The capacities of the load and transformer are set to the maximum of 10,000 (kVA) and 5000 (kVA), respectively. Fault currents flow through SW1 and SW2 according to the fault conditions, and the occurrence of FI is stored to be used as training and test data.
Table 2 and Table 3 show the training and test data, respectively, when a fault occurs in the data generation model, which is shown in Figure 9. In the FI validation results, ‘1’ represents valid FIs activated by the current flowing from the power side of the fault location, and ‘0’ means no/false FIs that are not generated or false FIs that are generated due to reverse fault currents.
After training the ANN using over 190,000 data, the results are tested and verified using the test data and validation data, respectively, as shown in Table 4. The batch and epoch for the ANN model are selected as 20 and 50, respectively, since those guarantee high performance in terms of accuracy and precision.
4. Case Studies
4.1. Case Descriptions
To verify the performance of the proposed method, real fault conditions, with the current data measured on-site for a 22.9 kV Korean distribution network causing incorrect fault section identification results, are considered. The data on the fault currents can be used to verify the performance of the proposed method since the ANN model requires only data on fault currents (A, B, C, and N phases) to validate FIs and the resulting fault section identifications.
Figure 10 shows the incorrect fault section identification cases with actual FIs generated by the reverse current flowing from the load side to the power side at the fault location due to the interconnection of DERs. In the case of Figure 10a, although the fault occurs between SW3, SW4, and SW8, the FIs of SW4, SW5, and SW10 are also activated due to the reverse current, resulting in the fault section being erroneously determined as between SW10 and DG3. Similarly, the FIs are generated at SW1, SW2, SW3, and SW4 even if the fault occurs in the section between CB and SW1, causing the fault section to be incorrectly identified as the section between SW4 and SW5. Table 5 shows the fault current data acquired from the system under the fault cases considered in Figure 10.
4.2. Results and Discussions
To verify the performance of the proposed method, real fault conditions, with the current data measured on-site causing incorrect fault section identification results, are considered.
Figure 11 represents the proposed ANN model results for the test data. The validation results of the FIs are represented by blue and red with 1 and 0, respectively. As shown in Figure 11, there are more false FIs as the zero-sequence current value increases.
Table 6 represents the ANN model results for the actual measured current values. Figure 12 shows that the fault section can be accurately identified using the results of the ANN model in Table 6. In the case of Figure 12a, the proposed ANN model determines that the FI generated at SW3 is caused by the current flowing from the power side to the load side, while the FIs activated at SW4, SW5, and SW10 are caused by the reverse currents. If the proposed fault section identification method is applied, it can be confirmed that the fault section is correctly identified as the section between SW3, SW4, and SW8, which matches the actual fault location. In the case of Figure 12b, it is determined that the FIs at SW1, SW2, SW3, and SW4 are activated due to the reverse currents, and CB is considered as the fault section because a lock-out occurs between CB and SW1. It can be confirmed that the fault section and actual fault location are correctly matched. As long as the training of the ANN model is well-performed, it is expected that the performance of the proposed method is guaranteed.
5. Conclusions
The DAS was developed to improve power supply reliability in distribution systems. However, with the interconnection of DERs in distribution networks, reverse fault currents can occur due to the load side of wye-delta transformers, DERs, and heavy loads, making the conventional fault section identification method unable to be used. A new approach to fault section identification using the validation of FIs based on an ANN is proposed. The proposed ANN model utilizes training data to recognize the relationship between fault current data and FI validation results. Fault section identifications can be performed by using the proposed method even when reverse fault currents from the downstream side of the fault location are present. In addition, the proposed method can be applied to distribution systems with multiple DGs without any additional equipment and communication infrastructure.
Author Contributions
Conceptualization, M.-S.K. and J.-U.S.; methodology, M.-S.K.; software, D.-H.K.; validation, M.-S.K. and J.-G.A.; formal analysis, M.-S.K. and J.-U.S.; investigation, Y.-S.O. and S.-I.L.; writing—original draft preparation, M.-S.K.; writing—review and editing, Y.-S.O., S.-I.L. and J.-U.S.; visualization, D.-H.K.; supervision, M.-S.K.; project administration S.-I.L.; funding acquisition, M.-S.K. All authors have read and agreed to the published version of the manuscript.
Funding
This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korea government (MOTIE) (20225500000060, Operation System for AC/DC Hybrid Distribution Networks).
Institutional Review Board Statement
Not applicable.
Informed Consent Statement
Not applicable.
Data Availability Statement
Not applicable.
Conflicts of Interest
The authors declare no conflict of interest.
Appendix A
Table A1.
Parameters for the sample distribution network.
| Parameter | Value |
|---|---|
| Positive-sequence resistance of a line section | 0.182321 (Ω/km) |
| Zero-sequence resistance of a line section | 0.504942 (Ω/km) |
| Positive-sequence reactance of a line section | 0.001035 (Ω/km) |
| Zero-sequence reactance of a line section | 0.002647 (Ω/km) |
| Resistance of a transformer (primary) | 0.000001 (Ω) |
| Reactance of a transformer (primary) | 0.036100 (Ω) |
| Resistance of a transformer (secondary) | 0.000001 (Ω) |
| Reactance of a transformer (secondary) | 131.1025 (Ω) |
References
- Kezunovic, M. Smart Fault Location for Smart Grids. IEEE Trans. Smart Grid 2011, 2, 11–22. [Google Scholar] [CrossRef]
- Zidan, A.; El-Saadany, E.F. A Cooperative Multiagent Framework for Self-Healing Mechanisms in Distribution Systems. IEEE Trans. Smart Grid 2012, 3, 1525–1539. [Google Scholar] [CrossRef]
- Chow, M.-Y.; Taylor, L.; Chow, M.-S. Time of outage restoration analysis in distribution systems. IEEE Trans. Power Deliv. 1996, 11, 1652–1658. [Google Scholar] [CrossRef] [Green Version]
- Cho, N.; Bhat, R.; Lee, J.-H. Field experience of distribution automation system (DAS) in Korea electric power corporation (KEPCO). In Proceedings of the 2015 International Symposium on Smart Electric Distribution Systems and Technologies, Vienna, Austria, 8–11 September 2015; pp. 307–312. [Google Scholar] [CrossRef]
- Lim, I.-H.; Sidhu, T.S.; Choi, M.-S.; Lee, S.-J.; Hong, S.; Lim, S.-I.; Lee, S.-W. Design and Implementation of Multiagent-Based Distributed Restoration System in DAS. IEEE Trans. Power Deliv. 2013, 28, 585–593. [Google Scholar] [CrossRef]
- Lim, I.; Sidhu, T.S.; Choi, M.S.; Lee, S.J.; Ha, B.N. An Optimal Composition and Placement of Automatic Switches in DAS. IEEE Trans. Power Deliv. 2013, 28, 1474–1482. [Google Scholar] [CrossRef]
- An, H.S.; Choi, M.H.; Seong, I.; Hyeong, S.Y.; Bae, S.H.; Song, N.H. A study of interconnections between renewable energy source and 22.9kV-Y distribution power system in Korea. In Proceedings of the 2014 International Conference on Renewable Energy Research and Application (ICRERA), Milwaukee, WI, USA, 19–22 October 2014; pp. 147–152. [Google Scholar]
- Liao, Y. Generalized Fault-Location Methods for Overhead Electric Distribution Systems. IEEE Trans. Power Deliv. 2011, 26, 53–64. [Google Scholar] [CrossRef]
- Jia, K.; Thomas, D.; Sumner, M. A New Single-Ended Fault-Location Scheme for Utilization in an Integrated Power System. IEEE Trans. Power Deliv. 2013, 28, 38–46. [Google Scholar] [CrossRef]
- Dashti, R.; Daisy, M.; Shaker, H.R.; Tahavori, M. Impedance-Based Fault Location Method for Four-Wire Power Distribution Networks. IEEE Access 2017, 6, 1342–1349. [Google Scholar] [CrossRef]
- Jia, K.; Bi, T.; Ren, Z.; Thomas, D.W.P.; Sumner, M. High Frequency Impedance Based Fault Location in Distribution System With DGs. IEEE Trans. Smart Grid 2018, 9, 807–816. [Google Scholar] [CrossRef]
- Li, Z.; Yao, J.; Zeng, X.; Deng, F. Power grid fault traveling wave network protection scheme. Int. J. Electr. Power Energy Syst. 2011, 33, 875–879. [Google Scholar]
- He, Z.; Zhang, J.; Li, W.-H.; Lin, X. Improved Fault-Location System for Railway Distribution System Using Superimposed Signal. IEEE Trans. Power Deliv. 2010, 25, 1899–1911. [Google Scholar] [CrossRef]
- Teng, J.-H.; Huang, W.-H.; Luan, S.-W. Automatic and Fast Faulted Line-Section Location Method for Distribution Systems Based on Fault Indicators. IEEE Trans. Power Syst. 2014, 29, 1653–1662. [Google Scholar] [CrossRef]
- Zhang, N.; Kezunovic, M. Transmission Line Boundary Protection Using Wavelet Transform and Neural Network. IEEE Trans. Power Deliv. 2007, 22, 859–869. [Google Scholar] [CrossRef]
- Chang, C.; Kumar, S.; Liu, B.; Khambadkone, A. Real-time detection using wavelet transform and neural network of short-circuit faults within a train in DC transit systems. IEE Proc. Electr. Power Appl. 2001, 148, 251–256. [Google Scholar] [CrossRef]
- Zhou, Y.; Kumar, A.; Parkash, C.; Vashishtha, G.; Tang, H.; Xiang, J. A novel entropy-based sparsity measure for prognosis of bearing defects and development of a sparsogram to select sensitive filtering band of an axial piston pump. Measurement 2022, 203, 111997. [Google Scholar] [CrossRef]
- Farooq, M.U.; Anwar, S.; Bhatti, H.A.; Kumar, M.S.; Ali, M.A.; Ammarullah, M.I. Electric Discharge Machining of Ti6Al4V ELI in Biomedical Industry: Parametric Analysis of Surface Functionalization and Tribological Characterization. Materials 2023, 16, 4458. [Google Scholar] [CrossRef]
- Jamari, J.; Ammarullah, M.I.; Santoso, G.; Sugiharto, S.; Supriyono, T.; Permana, M.S.; Winarni, T.I.; van der Heide, E. Adopted walking condition for computational simulation approach on bearing of hip joint prosthesis: Review over the past 30 years. Heliyon 2022, 8, e12050. [Google Scholar] [CrossRef]
- Prakoso, A.T.; Basri, H.; Adanta, D.; Yani, I.; Ammarullah, M.I.; Akbar, I.; Ghazali, F.A.; Syahrom, A.; Kamarul, T. The Effect of Tortuosity on Permeability of Porous Scaffold. Biomedicines 2023, 11, 427. [Google Scholar] [CrossRef]
- Ammarullah, M.I.; Santoso, G.; Sugiharto, S.; Supriyono, T.; Kurdi, O.; Tauviqirrahman, M.; Winarni, T.I.; Jamari, J. Tresca stress study of CoCrMo-on-CoCrMo bearings based on body mass index using 2D computational model. J. Tribol. 2022, 33, 31–38. [Google Scholar]
- Ammarullah, M.I.; Hartono, R.; Supriyono, T.; Santoso, G.; Sugiharto, S.; Permana, M.S. Polycrystalline Diamond as a Potential Material for the Hard-on-Hard Bearing of Total Hip Prosthesis: Von Mises Stress Analysis. Biomedicines 2023, 11, 951. [Google Scholar] [CrossRef]
- Ammarullah, M.I.; Santoso, G.; Sugiharto, S.; Supriyono, T.; Wibowo, D.B.; Kurdi, O.; Tauviqirrahman, M.; Jamari, J. Minimizing Risk of Failure from Ceramic-on-Ceramic Total Hip Prosthesis by Selecting Ceramic Materials Based on Tresca Stress. Sustainability 2022, 14, 13413. [Google Scholar] [CrossRef]
- Ammarullah, M.I.; Afif, I.Y.; Maula, M.I.; Winarni, T.I.; Tauviqirrahman, M.; Akbar, I.; Basri, H.; van der Heide, E.; Jamari, J. Tresca Stress Simulation of Metal-on-Metal Total Hip Arthroplasty during Normal Walking Activity. Materials 2021, 14, 7554. [Google Scholar] [CrossRef]
- Tauviqirrahman, M.; Jamari, J.; Susilowati, S.; Pujiastuti, C.; Setiyana, B.; Pasaribu, A.H.; Ammarullah, M.I. Performance Comparison of Newtonian and Non-Newtonian Fluid on a Heterogeneous Slip/No-Slip Journal Bearing System Based on CFD-FSI Method. Fluids 2022, 7, 225. [Google Scholar] [CrossRef]
- Lamura, M.D.P.; Hidayat, T.; Ammarullah, M.I.; Bayuseno, A.P.; Jamari, J. Study of contact mechanics between two brass solids in various diameter ratios and friction coefficient. Proc. Inst. Mech. Eng. Part J J. Eng. Tribol. 2023. [Google Scholar] [CrossRef]
Figure 1.
Conventional fault section identification method.
Figure 2.
Zero sequence circuit for a single line to ground fault.
Figure 3.
Vector diagram of phase fault currents.
Figure 4.
A case of incorrect identification using the conventional method.
Figure 5.
Distribution of fault current data.
Figure 6.
DAS equipped with the proposed ANN model.
Figure 7.
Flowchart for creating the learning model.
Figure 8.
The structure of the proposed ANN model.
Figure 9.
The sample distribution network for data generation.
Figure 10.
Fault cases with incorrect fault section identification. (a) Fault in the section between SW3, SW4, and SW8. (b) Fault in the section between CB and SW1.
Figure 10.
Fault cases with incorrect fault section identification. (a) Fault in the section between SW3, SW4, and SW8. (b) Fault in the section between CB and SW1.
Figure 11.
Results for the test data of the proposed ANN model.
Figure 12.
Results of fault section identification using the proposed ANN model. (a) Fault in the section between SW3, SW4, and SW8. (b) Fault in the section between CB and SW1.
Figure 12.
Results of fault section identification using the proposed ANN model. (a) Fault in the section between SW3, SW4, and SW8. (b) Fault in the section between CB and SW1.
Table 1.
Active power and class of loads in the sample system.
| Parameters | Range |
|---|---|
| Fault resistance (ohms) | 0–30 |
| Fault distance (km) | 0–30 |
| Load capacity (kVA) | 0–10,000 |
| Transformer capacity (kVA) | 0–5000 |
Table 2.
Training data generated by MATLAB.
| Index | Phase Fault Current (A) | FI Validation | |||
|---|---|---|---|---|---|
| A | B | C | N | ||
| 1 | 7455 | 0.013 | 0.013 | 7455 | 1 |
| 2 | 7459 | 68.51 | 65.05 | 7326 | 1 |
| 3 | 7463 | 134.8 | 127.8 | 7202 | 1 |
| 4 | 7472 | 261.0 | 247.0 | 6965 | 1 |
| 5 | 7493 | 596.2 | 560.8 | 6340 | 1 |
| 6 | 7522 | 1043 | 972.0 | 5514 | 1 |
| 7 | 7455 | 25.45 | 25.77 | 7442 | 1 |
| 8 | 7460 | 58.46 | 90.35 | 7314 | 1 |
| 9 | 7464 | 122.4 | 152.8 | 7190 | 1 |
| 10 | 7472 | 247.4 | 271.6 | 6954 | 1 |
| … | … | … | … | … | … |
| 192,706 | 220.4 | 124.1 | 82.43 | 247.0 | 0 |
| 192,707 | 233.6 | 142.6 | 61.67 | 298.8 | 0 |
Table 3.
Test data generated by MATLAB.
| Index | Phase Fault Current (A) | FI Validation | |||
|---|---|---|---|---|---|
| A | B | C | N | ||
| 1 | 2845 | 119.3 | 109.6 | 2628 | 1 |
| 2 | 2808 | 454.2 | 404.3 | 1985 | 1 |
| 3 | 2851 | 120.8 | 133.3 | 2608 | 1 |
| 4 | 2815 | 453.0 | 423.4 | 1973 | 1 |
| 5 | 2866 | 147.5 | 193.3 | 2558 | 1 |
| 6 | 2831 | 455.9 | 472.2 | 1942 | 1 |
| 7 | 2887 | 214.2 | 276.9 | 2490 | 1 |
| 8 | 2854 | 473.1 | 541.9 | 1899 | 1 |
| 9 | 2153 | 138.2 | 128.9 | 1904 | 1 |
| 10 | 2193 | 483.8 | 434.6 | 1319 | 1 |
| … | … | … | … | … | … |
| 44,866 | 238.4 | 156.7 | 44.35 | 325.9 | 0 |
| 44,867 | 238.7 | 180.8 | 10.47 | 356.0 | 0 |
Table 4.
Test results of the ANN model.
| Batch | Epoch | Accuracy | Precision | Recall | F1 Score |
|---|---|---|---|---|---|
| 20 | 10 | 0.9998 | 0.9996 | 1.0000 | 0.9998 |
| 50 | 1.0000 | 1.0000 | 1.0000 | 1.0000 | |
| 100 | 1.0000 | 1.0000 | 1.0000 | 1.0000 | |
| 100 | 10 | 0.9976 | 0.9952 | 1.0000 | 0.9976 |
| 50 | 1.0000 | 1.0000 | 1.0000 | 1.0000 | |
| 100 | 1.0000 | 1.0000 | 1.0000 | 1.0000 | |
| 500 | 10 | 0.9779 | 0.9575 | 1.0000 | 0.9783 |
| 50 | 0.9986 | 0.9972 | 1.0000 | 0.9986 | |
| 100 | 0.9998 | 0.9996 | 1.0000 | 0.9998 | |
| 1000 | 10 | 0.9356 | 0.8857 | 1.0000 | 0.9394 |
| 50 | 0.9988 | 0.9977 | 1.0000 | 0.9988 | |
| 100 | 1.0000 | 1.0000 | 1.0000 | 1.0000 |
Table 5.
On-site fault current data.
| Index | Phase Fault Current (A) | |||
|---|---|---|---|---|
| A | B | C | N | |
| 1 | 1156 | 151 | 141 | 1141 |
| 2 | 170 | 226 | 254 | 396 |
| 3 | 214 | 200 | 137 | 363 |
| 4 | 125 | 123 | 123 | 374 |
| 5 | 163 | 205 | 220 | 174 |
| 6 | 183 | 99 | 137 | 133 |
| … | … | … | … | … |
| 20 | 173 | 169 | 92 | 316 |
| 21 | 89 | 87 | 48 | 217 |
Table 6.
Test data generated by MATLAB for case studies.
| Index | Phase Fault Current (A) | FI Validation | |||
|---|---|---|---|---|---|
| A | B | C | N | ||
| 1 | 1156 | 151 | 141 | 1141 | 1 |
| 2 | 170 | 226 | 254 | 396 | 0 |
| 3 | 214 | 200 | 137 | 363 | 0 |
| 4 | 125 | 123 | 123 | 374 | 0 |
| 5 | 163 | 205 | 220 | 174 | 0 |
| 6 | 183 | 99 | 137 | 133 | … |
| … | … | … | … | … | … |
| 20 | 173 | 169 | 92 | 316 | 0 |
| 21 | 89 | 87 | 48 | 217 | 0 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
MDPI and ACS Style
Kim, M.-S.; An, J.-G.; Oh, Y.-S.; Lim, S.-I.; Kwak, D.-H.; Song, J.-U. A Method for Fault Section Identification of Distribution Networks Based on Validation of Fault Indicators Using Artificial Neural Network. Energies 2023, 16, 5397. https://doi.org/10.3390/en16145397
AMA Style
Kim M-S, An J-G, Oh Y-S, Lim S-I, Kwak D-H, Song J-U. A Method for Fault Section Identification of Distribution Networks Based on Validation of Fault Indicators Using Artificial Neural Network. Energies. 2023; 16(14):5397. https://doi.org/10.3390/en16145397
Chicago/Turabian StyleKim, Myong-Soo, Jae-Guk An, Yun-Sik Oh, Seong-Il Lim, Dong-Hee Kwak, and Jin-Uk Song. 2023. "A Method for Fault Section Identification of Distribution Networks Based on Validation of Fault Indicators Using Artificial Neural Network" Energies 16, no. 14: 5397. https://doi.org/10.3390/en16145397
Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.
