Title: Single Current Sensor Technique for Detecting, Classifying, and Localizing PV String Faults.
A
MuhammadAbdullah1✉Emailmqasim@bournemouth.ac.uk
MuhammadQasimShah2
FajarKabeer3
1Institute of Electrical, Electronics and Computer EngineeringUniversity of the PunjabLahorePakistan
2Design and Engineering, Faculty of Science and TechnologyBournemouth UniversityEnglandUK
3Department of PhysicsGovernment College UniversityLahorePakistan
1Muhammad Abdullah, Muhammad Qasim Shah2, Fajar Kabeer3.
1Institute of Electrical, Electronics and Computer Engineering, University of the Punjab, Lahore, Pakistan.
2Design and Engineering, Faculty of Science and Technology,Bournemouth University, England, UK,
3Department of Physics, Government College University, Lahore, Pakistan,
Abstract
In a photovoltaic (PV) system, faults such as Line to Ground (L_G) and Line to Line (L_L) can cause considerable efficiency losses and increase fire risks. Therefore, early detection, classification, and localization of these faults are essential for ensuring efficiency and safety. This paper introduces a new fault detection, classification, and localization algorithm. The algorithm uses information from a single current sensor positioned above the blocking diode in each PV string of the PV array. The key advantage of this algorithm is that it can operate effectively even in the presence of MPPT and blocking diodes.
Keywords:
(PV) system
Line to Ground (L_G)
Line to Line (L_L)
PV array
Corresponding author
mqasim@bournemouth.ac.uk
A
1. Introduction
In the recent decade, there has been a notable increase in the installation of photovoltaic (PV) systems. As this system is easy to maintain and install, it can be considered a noise-free alternative to the conventional power system.
Just like any other power system, the PV system is susceptible to faults like arc fault, line-to-line fault (L_L), line-to-ground fault (L_G), open circuit, and partial shading. In the large-scale PV system, the major cause of fires is related to the given faults. These faults can occur due to improper or loose connections, corrosive cells, and degradation of the internal and external elements of a PV panel. Sometimes these faults can occur due to the unbalanced currents [1–4].
According to several international safety standards like IEC 62109-1, IEC60269-6, IEC 62109-2, and NEC- 690, the PV system should be placed with some protection devices known as over current protection devices (OCPD), arc fault circuit interrupter (AFCI) and ground fault protection devices (GFPD) [5–7].
These devices are conceptualized and designed according to the PV array's open circuit voltages and short circuit current ratings. They are designed to incorporate the standard testing conditions parameters (STC) of a PV module. However, the PV system is highly non-linear, exhibiting varying performance and voltage and current values.
The protection devices are present, but the highly non-linear behaviors of a PV system during the day can hinder the detection of fault by (OCPD), (GFPD), and (AFCI) devices. Another cause can be the presence of bypass diodes and degrading insulation resistance due to all the discussed reasons; the conventional protection devices cannot cover the complete range of faults [7].
For instance, during a rainy season, a GFPD device rated 20A cannot detect fault current and trip to prevent potential harm, such as fire, because the fault current's value is uncertain [7].
The output current-voltage (I-V) curve of a photovoltaic array under the L_G and L_L fault is approximately identical to the (I-V) curve of a PV array experiencing a partial shading condition (PSC) [8–9], the situation becomes more drastic when there are blocking diodes present in the PV array the function of blocking diode is to not allow the reverse current to enter in a PV string, due to this the OCPD device will not be able to detect any kind of fault [10].
Given the challenges in accurately detecting and localizing L_G and L_L faults in PV systems, numerous researchers have focused on developing effective circuit and algorithmic systems for fault detection, classification, and localization. To detect, classify and localize the L_G and L_L fault accurately many researchers have been using sensory based data in which they use different kinds of sensors like current sensor and voltage sensors, In, [11] a technique is proposed which incorporates a voltage sensor to calculate the differential voltage which will detect and localize L_G, L_L fault at PV string level. In [10, 12, 13], voltage sensors and current sensors are used to detect, classify, and localize L_G and L_L faults accurately; however, the proposed methods require extra circuitry and a large number of sensors. In [13], two current sensors per string are used to detect and classify the L_G and L_L fault at string level, while the researchers in [12] incorporate three voltage sensors, each of which will have its independent function: one detects the fault, the second classifies the type, and the third will localize the fault.
Authors in [14] used a honeycomb and a bridge-like configuration by attaching voltage sensors across the top and bottom of the shorted terminal nodes with back-to-back strings to use the voltage measurements for the detection and location of faults. In continuation of the work cited in [14], sensors are placed to measure the magnitude of the voltage. These sensors are placed in locations where they can easily detect and localize faults, but they are also exposed to potential damage [15].
In the recent age of machine learning (ML) and artificial intelligence (AI), many researchers have also created algorithms to detect and classify faults in a PV array with the help of machine learning. Authors in [16] developed a complex artificial intelligence algorithm capable of detecting faults, but it is limited to bypass diodes and is also complex to understand. In [17–18], a compound scheme based on a neural network and the high-frequency sub-bands of current is used to detect L_G/L_L faults. However, its architecture is quite complex and requires specialized knowledge.
In this paper, a new algorithm for fault detection, classification, and localization based on a single current sensor per string is proposed. The details of the algorithm, sensor placement, and analysis are elaborated in sections 2 and 3.
2. Fault Analysis in a Photovoltaic Array
As discussed in the introduction, just like any other electrical system, a PV system is susceptible to faults. The types of faults are elaborated in 2.1 below,
2.1. Types of Faults
There are four major types of faults in a PV system: Open circuit fault, short circuit fault, either Line-Line (L_L), or Line-Ground. Now, the L_L and L_G faults are significant as they can drastically drop the performance of any PV system. These faults can be shown in Fig. 1.
Fig. 1
Representation of types of Faults.
2.2. Fault detection challenges
To protect the PV array from these faults, OCPD and GFPD were used, but due to various conditions like the placement of blocking diodes and moving clouds, these faults remain hidden from these devices [10]. Other challenges associated with these are already mentioned in the introduction section of this manuscript.
There are some challenges associated with the accurate detection of L_G and L_L faults, as the I-V curve of these faults is quite similar to that of shading conditions, and it remains unnoticed that L_G and L_L faults can cause a significant drop in efficiency and can cause a fire hazard.
3. Proposed Methodology
A PV array consists of PV strings connected in parallel, with the PV modules linked in series to form each PV string. The number of modules and strings varies depending on the application. Let the number of strings be S1, S2, S3, S4, ..., SN.
This manuscript presents a new algorithm that uses a single current sensor to detect, classify, and localize L_G and L_L faults in a PV string. In this method, one current sensor is installed above the blocking diode of each PV string in the array.
The algorithm relies on differences in current readings between two consecutive sensors; the current sensor detects this change and sends a command to the controller block attached to it, which is positioned above a blocking diode in each string. This new proposed circuit setup is shown in Fig. 2.
Fig. 2
Proposed Circuit Diagram.
In Fig. 2, a red-colored block is visible, which is primarily the control block responsible for detecting, classifying, and localizing L_G and L_L faults. This control block has four inputs—S1_U, S2_U, S3_U, and S4_U—that represent the current sensors installed in strings 1, 2, 3, and 4.
The sensor information coming from these sensors is then processed in this control block, and the result is the detection, classification, and localization of faults mentioned before. The process involves different mathematical equations for each type of fault.
3.1. Proposed Circuit Analysis of the L_G and L_L Faults
Fig. 3
The proposed circuit for visualizing the L_G fault.
In Fig. 3, the representation of the proposed circuit analysis based on a single current sensor can be seen. Firstly, the string operation region is set to be in the third region with the use of the MPPT block. When a Fault occurred in string 1 (S1), a fault current IF is produced, which will follow the less resistive path that is towards ground. The total string current, which was previously IMPP, will now become IS1_U. Due to the opposite direction of the IF value, it will be subtracted from IMPP, which is the current of a PV array in normal conditions.
Fig. 4
The proposed circuit for visualizing the L_L fault.
In Fig. 4, the diagram of the proposed circuit analysis using a single current sensor is shown. First, the string operation region is set to the third region via the MPPT block. When a fault occurs in string 1 (S1) and string 2 (S2), a fault current IF is generated. The total string current, which was previously IMPP, will now become IS1_U. Because of the opposite direction of the IF value, it will be subtracted from IMPP, which represents the current of a PV array under normal conditions. Now, the main difference from L_G and L_L is that the value of IF will vary depending on each fault scenario.
Let us consider the current through the sensor, represented by ISN_U in each string. For instance, if an L_G fault occurs, the control block will process Eq. 1.…………………..
1
As mentioned in Eq. (1), the 𝑆𝑛+1_𝑈− 𝑆𝑛_𝑈 represents the difference between two sensors, as discussed earlier in this section, and this difference is then processed in the control block. Initially, it classifies the fault as L_G and then determines the number of PV strings where the fault has occurred. Similarly, for the L_L fault, Eq. 2 will be used, which is as follows.…………………...
2
Just like Eq. (1), Eq. (2) shows a difference between sensor information; its representation changes slightly because it requires two different data sets—one fault point is in one string, and the other point is in the following string. The control block will check the fault prompt and classify and localize the faulty strings.
4. Results and Discussion
4.1. Proposed Algorithm
Figure 5 depicts the proposed algorithm, demonstrating how the control block shown in Fig. 2 functions to identify, classify, and localize L_G and L_L faults in a PV string.
Fig. 5
Flowchart of the Proposed Algorithm.
As previously mentioned in the materials and methods section, the PV array will initially be set to the third region of operation. The operation of this algorithm will be described below.
1.
The PV array is initially set to the third region of operation, which is 2Voc-3Voc in our case.
2.
The proposed algorithm will begin monitoring the current sensor values that are installed on top of the blocking diode in each string.
3.
Subsequently, the proposed algorithm will calculate the IF value, which essentially indicates the current that occurs only when the PV string experiences L_G or L_L faults. Then, a preset α value will be added to the IF. This preset value differs for L_G and L_L faults.
4.
After the previous step in the control block, the coding algorithm will determine the
and
.
5.
The final stage of this fault detection, classification, and localization process involves a series of conditional prompts. The program checks the conditions that allow it not only to detect L_G or L_L faults but also to classify the fault type and identify the string where it occurs, thus completing localization. The conditions are outlined in the red colored decision blocks within the algorithm's flowchart.
4.2. Simulation Validation
The discussion in the materials and methods section, along with the proposed algorithm, shows the validation of the new method for identifying, classifying, and localizing L_G and L_L faults in a PV array through simulation. The simulation is performed using PSIM version 2022. The simulated diagram is displayed in Fig. 6. This diagram is used to validate the proposed algorithm.
Fig. 6
Circuit diagram of the simulation in the PSIM environment.
4.2.1. Performance of the proposed algorithm in L_G fault in S1.
When an L_G fault occurs in S1, as shown in Fig. 3, it can also be observed when the L_G fault happens in the PV string displayed in Fig. 6. A fault current will be generated, flowing in the opposite direction of IMPP. In our case, IMPP is 3.45A, and after the fault, the current sensor will record IS1_U, as shown in Fig. 3, with its current value dropping to 2.85A, as shown in Fig. 7. The algorithm will perform this process using Eq. 1. It will detect this change of 0.60A which is the value of IF and then add α to it. Afterward, the program will compare the S1_U current sensor value and the S2_U current sensor value, and then calculate ∆𝐿_𝐺1. The result of this process can be seen in Fig. 8. The Preset value α is 0.05.
Fig. 7
S1_U current sensor in normal no-fault condition and under L_G fault.
Fig. 8
Output Value of the control block in L_G S1.
4.2.2. Performance of the proposed algorithm in L_G fault in S2.
Every process discussed in 4.2.1 will be followed here as well; the only difference will lie in the comparison of sensors. When an L_G fault occurs in S2, the sensor value comparison will now be between S2_U and S3_U. The fault identification and classification process will remain the same, but localization will differ as the program will calculate ∆𝐿_𝐺2. Results can be seen in Fig. 9.
Fig. 9
Output Value of the control block in L_G S2
4.2.3. Performance of the proposed algorithm in L_G fault in S3.
As discussed in the previous section, the proposed algorithm will check the conditional prompt related to the L_G fault. When an L_G fault occurs, it can be detected and classified by comparing sensors S4_U and S3_U. The localization will be done by incorporating ∆𝐿_𝐺3, and the result is shown in Fig. 10.
Fig. 10
Output Value of the control block in L_G S3
4.2.4. Performance of the proposed algorithm in L_G fault in S4.
As discussed in the previous sections, the proposed algorithm will check the conditional prompt related to the L_G fault. When an L_G fault occurs in S4, it can be detected and classified by comparing sensors S3_U and S4_U. The localization will be done by incorporating ∆𝐿_𝐺4, and the result is shown in Fig. 11.
Fig. 11
Output Value of the control block in L_G S4
4.2.5. Performance of the proposed algorithm in L_L fault among S1-S2
When an L_L fault occurs between S1 and S2, a fault current IF is generated. This fault current will differ from the fault current that happened during the L_G fault in the S1 scenario. The new value of the current sensor will now be IS1_U which is 2.14A, which can be calculated by subtracting it from the IMPP, which in our case is 3.45A, and the IF which is 1.31A in the L_L fault case This can be seen from Fig. 12. This difference of 1.31 is detected by the algorithm after this preset value α will be added to it which is 0.05 in our case.
Fig. 12
S1_U current sensor in normal no-fault condition and under L_L fault.
The output of the control block, responsible for detection, classification, and localization results, can be seen in Fig. 13. After the calculation of
.
Fig. 13
Output Value of the control block in L_L S1_S2
4.2.6. Performance of the proposed algorithm in L_L fault among S2-S3
When an L_L fault occurs in S2 and S3, the proposed algorithm checks the difference in the S2_U value by calculating the IF. Then, the algorithm calculates ∆𝐿_𝐿2,3, which allows it to easily classify and localize the L_L fault in S2 and S3, as illustrated in Fig. 14.
Fig. 14
Output Value of the control block in L_L S2_S3
4.2.7. Performance of the proposed algorithm in L_L fault among S3-S4
When an L_L fault occurs in S3 and S4, the proposed algorithm checks the difference in the S3_U value by calculating the IF. Then, the algorithm calculates ∆𝐿_𝐿3,4, which allows it to easily classify and localize the L_L fault in S2 and S3, as illustrated in Fig. 15.
Fig. 15
Output Value of the control block in L_L S3_S4
5. Conclusion
The proposed new algorithm is capable of detection, classification, and localization of L_G and L_L faults. The algorithm's functionality is validated through a series of simulations, which successfully detect L_G and L_L faults and classify them. Localization is limited to the PV string level, integrating both the MPPT block and blocking diodes. In the future, the work will be validated through hardware implementation and localization at the PV module level. Some versions of Machine learning can also be used, such as a decision tree.
A
A
Author Contribution
Abdullah/Fajar = Computational workQasim = Writeup
A
Acknowledgement
This research is novel, please consider it for publication.
REFERENCES:
1.
Y. Zhao, J.-F. de Palma, J. Mosesian, R. Lyons, and B. Lehman, “Line–line fault analysis and protection challenges in solar photovoltaic arrays,” IEEE Trans. Ind. Electron., vol. 60, no. 9, pp. 3784–3795, Sep. 2013.
2.
L. Xu, Z. Pan, C. Liang, and M. Lu, “A fault diagnosis method for PV arrays based on new feature extraction and improved the fuzzy C-mean clustering,” IEEE J. Photovolt., vol. 12, no. 3, pp. 833–843, May 2022.
3.
M. K. Alam, F. Khan, J. Johnson, and J. Flicker, “A comprehensive review of catastrophic faults in PV arrays: Types, detection, and mit igation techniques,” IEEE J. Photovolt., vol. 5, no. 3, pp. 982–997, May 2015.
4.
Y.-Y. Hong and R. A. Pula, “Methods of photovoltaic fault detection and classification: A review,” Energy Rep., vol. 8, pp. 5898–5929, Nov. 2022.
5.
W. Miao, Y. Luo, Y. Liu, F. Wang, F. Zhi, and X. Zhou, “Detection of line-to-ground and line-to-line faults based on fault voltage analysis in PV system,” in Proc. 7th Int. Conf. Power Renew. Energy (ICPRE), Sep. 2022, pp. 424–429.
6.
Y. Chaibi, M. Malvoni, A. Chouder, M. Boussetta, and M. Salhi, “Simple and efficient approach to detect and diagnose electrical faults and partial shading in photovoltaic systems,” Energy Convers. Manage., vol. 196, pp. 330–343, Sep. 2019.
7.
Y. Zhao, “Fault analysis in solar photovoltaic arrays,” Ph.D. dissertation, Dept. Elect. Comput. Eng., Northeastern Univ., Boston, MA, USA, 2011.
8.
D. S. Pillai and N. Rajasekar, “An MPPT-based sensorless line–line and line–ground fault detection technique for PV systems,” IEEE Trans. Power Electron., vol. 34, no. 9, pp. 8646–8659, Sep. 2019.
9.
A. Eskandari, J. Milimonfared, and M. Aghaei, “Fault detection and clas sification for photovoltaic systems based on hierarchical classification and machine learning technique,” IEEE Trans. Ind. Electron., vol. 68, no. 12, pp. 12750–12759, Dec. 2021.
10.
A. Mehmood, H. A. Sher, A. F. Murtaza, and K. Al-Haddad, “A diode-based fault detection, classification, and localization method for photovoltaic array,” IEEE Trans. Instrum. Meas., vol. 70, pp. 1–12, 2021.
11.
B. P. Kumar, D. S. Pillai, N. Rajasekar, M. Chakkarapani, and G. S. Ilango, “Identification and localization of array faults with opti mized placement of voltage sensors in a PV system,” IEEE Trans. Ind. Electron., vol. 68, no. 7, pp. 5921–5931, Jul. 2021.
12.
A. Mehmood, H. A. Sher, A. F. Murtaza, and K. Al-Haddad, “Fault detection, classification and localization algorithm for photovoltaic array,” IEEE Trans. Energy Convers., vol. 36, no. 4, pp. 2945–2955, Dec. 2021.
13.
A. F. Murtaza, M. Bilal, R. Ahmad, and H. A. Sher, “A circuit analysis based fault finding algorithm for photovoltaic array under L–L/L–G faults,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 8, no. 3, pp. 3067–3076, Sep. 2020.
14.
S.Ganesan,P.W.David,P.K.Balachandran,T.Senjyu,Faultidentificationscheme forsolarphotovoltaicarrayinbridgeandhoneycombconfiguration,Electr.Eng. 105(4) (2023)2443–2460.
15.
S. Ganesan, P.W. David, P.Murugesan, P.K. Balachandran, Solar photovoltaic systemperformance improvement using a new fault identification technique, Electr.PowerCompon. Syst.52(1) (2024)42–54.
16.
M. Dhimish and A. M. Tyrrell, “Photovoltaic bypass diode fault detec tion using artificial neural networks,” IEEE Trans. Instrum. Meas., vol. 72, pp. 1–10, 2023.
17.
Q. Liu, B. Yang, Y. Liu, K. Ma, and X. Guan, “Collaborate global and local: An efficient PV compound fault diagnosis scheme with multilabel learning and model fusion,” IEEE Trans. Instrum. Meas., vol. 72, pp. 1–16, 2023.
18.
S. A. Saleh, S. Kanukollu, and A. Al-Durra, “Phaselet transform based digital ground fault protection of grid-connected photovoltaic systems,” IEEE Trans. Ind. Appl., vol. 59, no. 5, pp. 5398–5410, Oct. 2023.