Review of Ground Insulation Resistance Detection Methods for Photovoltaic (PV) Systems

Renewable energy is becoming increasingly important in all sectors as the world moves toward a more sustainable energy future [1,2,3]. The PV industry, with its numerous benefits such as cleanliness and sustainability, has grown rapidly and emerged as a key player in the energy sector. The safe and stable operation of PV systems has a significant impact on power generation efficiency and reliability. However, the ground insulation resistance of PV modules is particularly vulnerable to environmental factors, which could pose a ground safety risk. Specifically, the performance of insulating materials gradually deteriorates due to multiple factors, such as aging, mechanical damage, electrical stress, and environmental erosion [4,5,6,7]. When the ground insulation resistance falls below a certain threshold, abnormal leakage currents may occur. This can jeopardize personal safety, cause equipment overheating [8,9], and undermine system reliability. Therefore, insulation resistance is a critical parameter that directly reflects the electrical insulation quality of PV systems and is required to ensure system reliability and energy efficiency. Consequently, developing effective detection methods is critical for ensuring the reliability of PV systems [10,11].

Numerous studies have investigated ground insulation resistance detection technologies for PV systems, proposing several detection methods [12,13,14,15,16,17,18,19]. These studies are primarily concerned with addressing the following core challenges.

In order to systematically summarize solutions to these challenges and provide practical references, this paper analyzes common detection methods by comparing their technical principles and performance. Furthermore, fault locating strategies at the system-, string-, branch-, and module-levels are discussed.

The structure of this article is as follows. Section 2 discusses the fundamental theory of ground insulation resistance in PV systems. Section 3 provides an in-depth review of various ground insulation resistance detection methods, evaluating their performance from technical principles to implementation mechanisms. Section 4 examines fault locating technologies at various levels and discusses the performance of different methods. Section 5 concludes the paper with a comprehensive summary and outlook on future developments.

Figure 1 depicts a typical PV system [1]. In this system, ground insulation resistance is classified into two types: DC-side and AC-side. On the DC-side, the ground insulation resistance is represented by R_p and R_n, which represent the resistance from the PV positive and negative poles to ground. On the AC-side, the ground insulation resistance is represented by R_a, R_b, and R_c, which represent the resistance of each AC phase to ground.

Higher insulation resistance indicates better insulation performance, which can protect against potential safety hazards caused by leakage currents. To ensure the safety of PV systems, ground insulation resistance should always be higher than the safety threshold. Therefore, the primary goal of insulation detection is to determine whether the ground insulation resistance falls below the safety threshold [5,6,7,8,9].

Because ground insulation resistance detection methods are inextricably linked to system architecture, the following introduces several common system architectures first.

Figure 2 depicts a string-type PV system with multiple PV strings connected in parallel to the DC-bus via DC-DC converters [20,21,22,23]. The ground resistance of PV strings is typically detected in two ways: (1) independently for each string, and (2) centralized monitoring via the DC-bus. It is important to point out that the ground insulation resistance detection and fault locating methods described in this paper mainly focus on string-type PV systems. However, these methods can also be applied to other types of PV systems.

Ground resistance detection is also closely linked to the grounding structure on the AC-side. There are two types of AC-side grounding structures: the terre-neutral (TN) system shown in Figure 3a and the isolated-terre (IT) system shown in Figure 3b,c [24,25,26,27].

In TN systems, the neutral point N′ on the AC-side is directly grounded. In these systems, if a ground fault occurs on the AC-side, the system’s protective devices will immediately disconnect the system, ensuring safety compliance. Therefore, insulation detection on the AC-side is not required in TN systems.

In IT systems, the neutral point N′ is either ungrounded or grounded via a high-resistance resistor. When a ground fault occurs, the fault current may remain below the protective threshold. However, the system continues to pose a risk in this scenario, necessitating insulation detection. As a result, the insulation resistance detection on the AC-side, discussed later in this paper, is primarily for IT systems.

When an insulation fault occurs, the source of the fault must be identified as soon as possible to ensure system safety. As shown in Figure 4, four levels of fault locating methods are commonly investigated: system-, string-, branch-, and module-levels.

This section divides insulation detection methods into two categories based on the need for additional hardware: hardware-based solutions that require additional hardware and software-based solutions that do not.

This method necessitates the integration of additional detection units into the system. There are three commonly used methods: the bridge detection unit method, the power injection method, and the self-creating ground circulation loop method.

The method includes a bridge detection circuit, which can be implemented using a resistive voltage divider network. The basic idea is to change the electrical parameters in the added network using internal switches, and then use multiple electrical equations to calculate the system’s ground insulation resistance [28,29,30,31]. Notably, this additional bridge detection circuit can be installed on either the PV- or DC-bus side.

Figure 5 depicts a detailed PV system with a single-switch topology detection circuit on the PV-side, while the equivalent circuit of this system is also presented [32]. As shown in Figure 5a, resistances R₁, R₂, and R₃ form a voltage divider network with an additional switch S_d. By turning the switch S_d on and off, two electrical equations can be obtained. Using these two equations, the system ground insulation resistance can be calculated.

From Figure 5b, when switch S_d is turned on, the voltage from ground to PV negative V_n(1) is obtained, as shown in Equation (1).

Similarly, when switch S_d is turned off, the voltage from PV negative to ground V_n(2) is obtained, as shown in Equation (2).

With Equation (1) and Equation (2), the ground insulation resistance (R_p and R_n) can be calculated.

Figure 6 shows alternative topologies for the detection circuit. Figure 6a depicts a topology that uses adjustable resistances rather than additional switch devices, which can reduce hardware costs [33]. Figure 6b depicts a dual-switch topology, which can enhance detection accuracy [34].

Papers [35,36] propose a method for detecting AC-side ground insulation resistance with the same bridge detection unit, as shown in Figure 7. The basic idea is to turn on one switch at a time on the bridge arms of phases A, B, and C, then connect the corresponding AC-side ground resistance (R_a, R_b, and R_c) in parallel with R_p or R_n. As shown in Figure 7a, when S₁ is turned on, R_a is connected in parallel with R_p. The equivalent circuit is presented in Figure 7b. From this, the AC-side insulation resistance R_a can be calculated using electrical equations. Similarly, by closing S₃ or S₅, the ground resistances of phases B and C (R_b and R_c) can also be calculated.

It should be noted that real-time calculation of mathematical equations is difficult; thus, some researchers have proposed several alternative methods to address this problem [37,38]. The basic idea is to measure the detection circuit’s voltage rather than calculating its resistance. When this voltage exceeds the predetermined threshold, it indicates the presence of an insulation fault. This method has been widely used in practice.

Although the bridge detection unit method is widely employed in PV systems, there are still several limitations.

Unlike the bridge detection unit method, this method’s detection circuit requires a voltage source and a current-limiting resistor, rather than resistor bridges. The basic idea behind this method is to perform the detection using a power signal generated by the added detection unit [39,40,41,42]. Several electrical equations are then formulated to calculate the system’s ground insulation resistance.

Papers [43,44] propose a technique that involves installing a detection unit between the PV negative and the ground. Figure 8 shows the hardware circuit and equivalent circuit implementing this technique in a PV system. As shown in Figure 8a, a voltage signal can be injected between the PV’s negative pole and ground. In fact, this voltage signal can also be injected into the DC-bus.

As shown in Figure 8b, to construct the electrical equations, initially set the voltage source V_T to V_T(1), where the subscript ₍₁₎ indicates the first measurement. Then, measure the voltage of PV negative to ground and get the value V_{PV_G(1)}. Then, Equation (3) can be established.

Similarly, for the second measurement, set the voltage source V_T to V_T(2), where the subscript ₍₂₎ indicates the second measurement. Then, measure the voltage of PV negative to ground and get the value V_{PV_G(2)}. Then, Equation (4) can be established.

With Equation (3) and Equation (4), the ground insulation resistance (R_p and R_n) can be established.

Similar to the bridge detection circuit method, the power injection method can assess insulation performance by measuring the voltage of the detection circuit [45].

Papers [46,47] propose a method for generating a voltage signal between the virtual neutral point of the AC-side N′ and ground, with three diodes used to create the virtual neutral point N′. Figure 9 depicts the hardware circuit and equivalent circuit implementing this method in a PV system.

In Figure 9b, set the voltage source as V_T. Then, measure the voltage across the current-limiting resistor R and get the value V_RT. Then, Equation (5) can be established.

where R is R_a//R_b//R_c.

It is worth noting that the above-described method can be used to calculate the total ground resistance of all three phases. To determine the ground resistance for each phase, detection circuits can be added separately between each phase on the AC-side and the ground [44].

Paper [43] proposes an alternative method to reduce system costs by sharing a common voltage source V_T between the DC- and AC-side. As illustrated in Figure 10, by controlling switches S₁ and S₂, the ground insulation resistance on the DC- and AC-side can be independently measured using the common voltage source.

It is important to note that the above-mentioned ground resistance detection method must be implemented prior to system startup; otherwise, the results may be inaccurate once operational. Therefore, papers [48,49] propose a method for measuring ground resistance while the system is operational. The topology of this system and its equivalent circuit are shown in Figure 11. From it, V_{dc_equ} and V_{ac_equ} are the equivalent DC and AC voltages, respectively, and R_s is the equivalent ground resistance.

Because the AC equivalent voltage (V_{ac_equ}) varies over time, detecting ground insulation resistance can be difficult. Therefore, a novel concept is proposed to cancel out the varying AC voltage [48,49]. Set two time points as t₁ and t₂, and adjust the voltage V_T to two different values as V_T(t1) and V_T(t2). Then, at these two time points, measure the corresponding current of I as I_(t1) and I_(t2), as well as the voltage of R_T as R_T(t1) and R_T(t2). Equation (6) and Equation (7) can be established.

In order to cancel out the AC voltage V_{ac_equ}, the time interval between t₁ and t₂ is set to an integer multiple of AC fundamental period, as shown in Figure 12. Then, V_{ac_equ(t1)} = V_{ac_equ(t2)}. By combining Equation (6) and Equation (7), R_s can be calculated as shown in Equation (8).

When R_s falls below the threshold, it indicates that the system ground resistance is low, causing the system to shut down. If the ground insulation resistance remains low, it is possible that an insulation fault occurred on the AC-side; otherwise, the fault is on the DC-side. This is because, when the system is turned off, the DC-side circuit is disconnected from the detection loop, making measuring the DC-side resistance impossible. Thus, only the AC-side ground resistance can be measured during this time.

The primary benefits of the power injection method are as follows.

However, practical limitations must be considered during implementation, particularly ensuring that the added detection unit does not interfere with normal operation. While the power injection method exhibits a negligible impact during offline execution, its online application requires strict injection signal constraints. Specifically, to reduce harmonic pollution and resonance risks, the signal frequency must not intersect with key grid harmonics or AC filter resonance points. Meanwhile, its amplitude must be limited to avoid false activation of overvoltage or overcurrent protection in the system.

Based on the fact that leakage current increases as system insulation deteriorates, this method uses leakage current to determine ground insulation status [40,50,51,52]. To create the leakage current loop, one typical approach is to sequentially switch on the power switches originally installed in the system [53]. Electrical equations can be established using the generated leakage current loop, allowing for the calculation of ground insulation resistance. Figure 13 illustrates the grounding circulating loops when different switches are turned on.

As shown in Figure 13a, when switch S₄ is turned on, ground circulating loops can be generated. It can be seen that the current flowing through R_p is I_Rp = V_pv/R_p, whereas the current flowing through R_n is theoretically zero. As a result, the leakage current is I_leak1 = −I_Rp, and Equation (9) can be established.

The method described above only detects the ground resistance R_p. To detect the ground resistance R_n, the upper switch S₃ can be turned on. As shown in Figure 13b, when switch S₃ is turned on, the current flowing through R_p is I_Rp = (V_pv − V_dc)/R_p, whereas the current flowing through R_n is I_Rn = V_dc/R_n. Consequently, the leakage current is I_leak2 = I_Rn − I_Rp, and Equation (10) can be established.

Furthermore, by combining Equation (9) and Equation (10), the actual values of R_p and R_n can be calculated.

It should be noted that if the grid neutral point is not grounded, the method described above cannot be applied. In this case, papers [54,55] propose a method that involves adding a detection unit between the AC- or DC-side and ground. Similarly, by turning on the appropriate switch, a ground circulating loop can be created. Then, the electrical equation is established to calculate the ground insulation resistance.

Figure 14 shows the added detection unit installed between the DC-side and ground. The added unit connects the PV positive or negative pole to ground via switches K₁ and K₂.

When K₁ is turned on, as shown in Figure 14a, the current I_Rn = V_pv/(R_n + R_T). When K₂ is turned on, as shown in Figure 14b, the current I_Rp = V_pv/(R_p + R_T). By combining these two equations, R_p and R_n can be calculated.

From above, this method is simple to implement and has a quick response time, allowing it to detect serious insulation faults such as ground short circuits. The primary technical challenge is that the system’s distributed capacitor can cause capacitive leakage current, which can be combined with resistive leakage current. This interference may reduce the accuracy of insulation detection.

In contrast to hardware-based solutions, software-based solutions do not require any additional detection units. Instead, system electrical parameters are monitored and observed to detect insulation resistance [56,57,58,59,60,61,62,63,64,65].

This method determines the insulation state by measuring the sum of currents at the positive and negative poles on the DC-side. When the PV system is operating normally, as shown in Figure 15, the sum of currents I_p and I_n is nearly zero [59,60]. However, if there is a ground fault on the DC-side, some current will flow into the ground, resulting in positive and negative imbalances and a significant increase in the current sum [61].

There are two limitations to this method: (1) It cannot determine whether the fault is located at the positive or negative pole. (2) It is ineffective when ground insulation faults occur at both the positive and negative poles.

This method enables ground insulation monitoring based on ground voltage on either the DC- or AC-side. The basic principle is that when a ground fault occurs, the ground voltage shifts, allowing for monitoring.

Figure 16 shows the ground voltage in a PV system. Ideally, the ground voltages across R_p and R_n should be equal, with V_pv+G = V_pv−G. However, when an insulation fault occurs, ground voltage shifting may occur. Specifically, as R_p decreases, V_pv+G decreases while V_pv−G increases, and vice versa. As a result, monitoring ground voltages can be used to determine whether an insulation fault exists at a specific PV pole [62].

Similarly, the ground voltages V_{A_G}, V_{B_G}, and V_{C_G} for AC phases A, B, and C should ideally be equal. When a ground fault occurs in one phase, the ground voltage of the faulty phase drops to almost zero, while the ground voltage of the non-faulty phase increases. Based on this feature, monitor the ground voltages of each phase in real time and compare them to the threshold. As a result, it can determine whether a ground fault has occurred on the AC-side [63].

According to the analysis above, one advantage of this method is its ability to identify whether faults occur at the positive or negative poles. However, this method is unable to detect symmetrical insulation faults, such as concurrent ground faults on both PV poles or all three AC phases. Furthermore, its performance is affected by system operating conditions. The AC-side load imbalance, in particular, reduces its sensitivity to early-stage faults, resulting in missed or false alarms.

In addition to monitoring insulation via current or ground voltage, these two parameters can be combined to calculate the ground insulation resistance in the system [64].

Figure 17 depicts an alternative solution for monitoring leakage current and DC-bus voltage. First, measure the ground voltage of the DC-bus negative V_{dc_G}, as well as the system’s leakage current I_leak. Then, V_{dc_G} and I_leak are filtered to remove high-frequency signals, yielding DC components V_{dc_G(DC)} and I_leak(DC). Therefore, the ground insulation resistance R_n is calculated with the equation of R_n = V_{dc_G(DC)} × K/I_leak(DC). In the equation, K is the adjustment factor, which is used to correct the shunt effect of R_p on the leakage current.

When calculating total ground insulation resistance on the AC-side, the three-phase power grid is equivalent to a voltage source. R represents the load between the voltage source to ground, and the current flowing through it is I_leak(DC). Because the positive pole of this voltage source is at the same potential as the midpoint of the DC-bus, the voltage from the positive pole to ground equals V_dc/2 + V_{dc_G(DC)}. Thus, the resistance R can be calculated as R = (V_dc/2 + V_{dc_G(DC)})/I_leak(DC).

Furthermore, this method can assess insulation status by tracking dynamic changes in V_{dc_G(DC)} and I_leak(DC). When these parameters exceed their associated thresholds, it indicates that the system has an insulation fault.

Since the system’s existing safety capacitors and ground insulation resistance form an RC current loop, the changes in insulation resistance can affect capacitor voltage. As a result, this method monitors the insulation status by measuring the voltage variation of the capacitor [65].

Figure 18 shows a PV system with a filter, where the safety capacitors C₁ and C₂ are installed between the positive and negative poles and the ground. Since C₁ and C₂ are connected in parallel with R_p and R_n physically, resulting in an RC loop.

When the insulation status is normal, both R_p and R_n are extremely high, resulting in a large time constant (τ = R × C) for the circuit loop. Under these conditions, the capacitor voltages V_C1 and V_C2 remain nearly constant between adjacent sampling intervals. However, if an insulation fault occurs, the decrease of R_p and R_n reduces the time constant τ, causing observable changes in V_C1 and V_C2 [65]. By monitoring the voltage values V_C1 and V_C2 in real time and assessing whether their variations between consecutive sampling intervals exceed a threshold, it is possible to detect changes in R_p and R_n and thus identify insulation faults.

It should be noted that this method only provides preliminary information on whether the ground insulation resistance has changed. Specific fault detection requires further evaluation in conjunction with other methods.

Table 1 presents the evaluation of different methods, including the dimensions of detection range, operating mode, impact on the system, limitations, and cost.

The table indicates that the self-creating ground circulation loop method, the current judgment method, and the capacitance voltage judgment method are primarily utilized for insulation detection on the DC-side, whereas the other methods can be applied to insulation detection on both the DC- and AC-side. Among these methods, the bridge detection unit method is primarily used for offline detection, the voltage and current joint-judgment method is used for online detection, while the others can be employed in both offline and online modes.

Furthermore, these methods exhibit certain limitations. Among hardware-based solutions, the bridge detection unit method poses challenges for real-time calculation. The power injection method limits the frequency and amplitude of the injected signal during online operation to avoid interference with normal system function. Furthermore, capacitive leakage current may have a negative impact on the effectiveness of the self-creating ground circulation loop method. It is important to note that all methods are influenced by the environment. Most methods rely on the PV array’s operating voltage, which severely limits their functionality at night or in low light conditions. Under these conditions, the power injection method with its own excitation source continues to be operational. However, all methods encounter difficulties during rapid irradiance transients, as the resulting voltage fluctuations can introduce significant measurement noise.

In terms of hardware costs, both the bridge detection unit method and the self-creating ground circulation loop method have low-cost, requiring only a few resistors and switches. The cost of the power injection method varies based on implementation. Using a dedicated detection circuit can be costly, but reusing the existing PID recovery equipment can significantly reduce the cost. Furthermore, all software-based solutions have extremely low hardware costs because they do not require additional hardware components.

In practical applications, factors such as system operating conditions, detection accuracy requirements, and cost should all be taken into account when determining the most effective detection method.

This section discusses insulation fault locating technologies at the system-, string-, branch-, and module-levels [66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83].

The system-level fault locating is designed to determine whether the fault is on the DC- or AC-side of the system.

Paper [73] proposed a system-level fault locating method, which analyzes frequency-domain characteristics of DC-bus ground voltages.

Figure 19 shows the ground voltage at the DC-bus in a PV system. In general, the DC-bus positive ground voltage V_dc+G and DC-bus negative ground voltage V_dc−G have a relatively large DC component, while the AC component is relatively small. When an insulation fault occurs on either DC-side pole, the corresponding ground voltage approaches 0. As a result, the DC component of the corresponding ground voltage is close to zero. Thus, the DC components of the DC-bus ground voltage can be used to identify DC-side faults.

If there is an insulation fault on the AC-side, the leakage current will couple to the DC-side, resulting in a significant AC component in the DC-bus ground voltage. Thus, the AC components of ground voltage can be used to identify an AC-side fault.

From above, Figure 20 depicts the whole process of system-level fault locating. The general idea is that if the DC component of the DC-bus ground voltage is less than the threshold, it is assumed that an insulation fault exists on the DC-side. If the AC component in the DC-bus ground voltage exceeds the threshold, the fault can be identified as being on the AC-side.

Alternatively, paper [74] proposes another fault locating method that involves measuring ground voltage at the DC-bus midpoint, V_{N_G}. Figure 21 illustrates the detailed fault locating process of this method.

In normal operation, the DC-bus midpoint N and the AC-side neutral point N′ remain at the same potential, and the ground voltage V_{N_G} is typically 0 V. However, if there is an insulation fault on DC-side, the ground voltages at the positive and negative poles are not equal, resulting in a voltage shift at DC-bus midpoint. If a ground fault occurs on the positive pole of the DC-bus, the ground voltage V_{N_G} is shifted to −V_dc/2. In contrast, when a ground fault occurs at the negative pole, the ground voltage V_{N_G} is shifted to +V_dc/2.

Furthermore, when one of the three AC phases experiences a ground fault, the ground voltage for that phase is zero. However, it does not affect the ground voltage of the DC-bus midpoint, V_{N_G}.

As a result, if the ground voltage at the midpoint deviates significantly from 0, an insulation fault on the DC-side is identified, which can be classified as a positive or negative pole fault based on the detailed voltage offset. Otherwise, it is determined that the insulation fault exists on the AC-side. The faulty phase is then identified by analyzing each phase’s ground voltage.

Figure 22 shows a PV system with multiple PV strings connected in parallel to the DC-bus. In order to achieve string-level fault locating, two typical locating methods are introduced: the string-by-string fault locating method and the perturbation and observation method.

Papers [45,75,76] propose a method for carrying out string-by-string fault locating. The key to this method is to connect only one PV string to the system at a time, isolating the others.

Figure 23 presents a detailed example. To check string PV₁, turn off the corresponding Boost switch Q₁. Also, turn on switches in the rest of the strings (Q₂, …, Q_n) simultaneously. As a result, the remaining PV strings (PV₂, …, PV_n) are separated from the system. Then, measure the system’s ground resistance, which is considered the ground resistance for string PV1. Similarly, other strings can be checked in the same way. Figure 24 depicts the entire procedure for all the strings. From above, this method has the advantage of accurately locating all insulation fault strings. However, the locating time may take a while.

This method involves applying voltage perturbations to each PV string and analyzing the system’s response to these perturbations to locate insulation faults.

Figure 25 depicts the voltage and current in a PV system with multiple PV strings. Figure 26 depicts the scenario in which a voltage perturbation is applied to PV₁. When the output voltage V_pv1 changes by ΔV, the ground voltage of the positive pole changes by ΔV_p = ΔV∗[R_p1/(R_p1 + R_n1)], while the ground voltage of the negative pole changes by ΔV_n = ΔV∗[R_n1/(R_p1 + R_n1)]. If string PV1 is properly insulated, the ground insulation resistances R_p1 and R_n1 are equivalent. Thus, ΔV_p equals ΔV_n. Otherwise, if there are insulation faults in string PV₁, the ground insulation resistances of the positive and negative poles will be different. This can result in significant differences in ΔV_p and ΔV_n [77].

In order to implement perturbations in practice, PV I-V scanning is normally applied together. As a result, the fault locating process begins by adjusting the operating point on the I-V curve, which causes a change in the PV string’s output voltage. Then, the difference between ΔV_p and ΔV_n is measured [78]. Repeat this process for each PV string. If the difference between ΔV_p and ΔV_n in a specific string exceeds the threshold, it indicates an insulation failure in this string.

Furthermore, paper [79] proposes a perturbing method, which is applied to each string’s ground voltage and then monitors the leakage current to find the insulation fault. When an insulation fault occurs in a PV string, a low-resistance path is formed between the fault point and ground. If the string’s ground voltage changes, additional fault current flows through the ground path, causing a significant change in system leakage current. As a result, measuring the system’s leakage current while varying the ground voltage can aid in locating the fault.

This section introduces branch-level fault locating. Taking string PV₁ in the PV system as an example, Figure 27 depicts the detailed branch structure. From it, the string PV₁ is made up of multiple PV branches PV₁₁ to PV_1n connected in parallel.

Papers [80,81,82] propose a fault locating method based on leakage current changes across branches. The principle is based on the idea that the value and direction of leakage current for faulty branches differ from normal conditions.

The following uses PV₁₁ ground fault as an example. Under normal conditions (see Figure 27a), all leakage currents flow in the same direction and have the same values for all branches.

In faulty conditions (see Figure 27b), the normal branches’ I_leak2~I_leakn direction is the same as in normal conditions. In contrast, the current I_leak1 of the faulty branch equals the sum of all other branch leakage currents, which is significantly greater than that in normal conditions. In addition, the current I_leak1 flows in the opposite direction than it would in normal operation.

Based on this, the faulty branch can be identified by examining the leakage current values and directions of each branch.

This section discusses module-level fault locating. Papers [78,83] propose a method for precisely locating the faulty module based on ground voltage distribution characteristics.

Figure 28 depicts the module-level structure of branch PV₁₁ in the PV system. When an insulation fault occurs in a PV module, its potential approaches ground potential, resulting in a significant change in the ground voltage distribution in the PV branch. The details are as follows.

From Figure 28, Equation (11) can be obtained.

Assume the #x module is faulty to the ground, the location #x can be determined with Equation (12).

It’s worth noting that the method described above is only effective for single-point faults in PV modules. Furthermore, external thermal imaging and electroluminescence imaging technologies can be used together to locate multiple-point faults [84].

Table 2 summarizes the characteristics of each locating method. System-level fault locating is simple to implement and inexpensive. The string-by-string fault locating method can be used to locate multiple faults, but it is slow. The perturbation and observation method is also easy to implement. Branch-level fault locating and module-level fault locating are only applicable to single-point faults.

Furthermore, in order to select an economical and efficient locating strategy in practical applications, several factors must be considered, including system scale, operating conditions, and maintenance needs.

This paper provides a systematic review of key technologies for detecting insulation resistance and locating faults in PV systems. Two types of insulation resistance detection methods are presented: hardware-based solutions and software-based solutions. The hardware-based solutions include the bridge detection unit method, power injection method, and self-creating ground circulation loop method. The software-based solutions are the current judgment method, ground voltage judgment method, voltage and current joint-judgment method, and capacitance voltage judgment method. For summarized methods, the detection range, operating mode, impact on the system, and limitations are all considered.

Following that, fault locating methods at the system-, string-, branch-, and module-levels are discussed. The system-level fault locating method determines whether the fault is on the DC- or AC-side by measuring ground voltage. String-level fault locating is divided into two methods: the string-by-string fault locating method and the perturbation and observation method. Branch-level fault locating is based on the value and direction of leakage current. The module-level fault locating method uses ground voltage distribution characteristics to identify the faulty PV module.

Future research can be conducted in the following directions. (1) Develop a low-cost online insulation detection method that meets the requirements for real-time performance, cost-effectiveness, and large-scale deployment in renewable energy scenarios. (2) Use artificial intelligence technology to reduce false alarms caused by noise interference and environmental changes, improving the accuracy of detection methods in complex scenarios. (3) Develop an intelligent locating method capable of collaborative fault locating at multiple levels and faults, allowing for quick and accurate fault isolation.

Generative AI was used solely for English grammar polishing.

The authors would like to thank the Technical Development Center of Ginlong Technologies Co., Ltd. for their technical assistance and access to the specialized database.

Conceptualization, W.Z. and L.X.; Methodology, W.Z. and L.X.; Formal Analysis, W.Z.; Investigation, W.Z. and L.X.; Resources, Y.W. (Yiming Wang 1) and P.X.; Data Curation, Y.W. (Yiming Wang 2); Writing—Original Draft Preparation, W.Z. and L.X.; Writing—Review & Editing, W.Z., L.X., Y.W. (Yiming Wang 1), Y.W. (Yiming Wang 2) and K.M.; Visualization, W.Z. and L.X.; Supervision, W.Z., P.X. and K.M.; Project Administration, Y.W. (Yiming Wang 1) and P.X.; Funding Acquisition, Y.W. (Yiming Wang 1).

Not applicable.

Not applicable.

Data availability is not applicable to this paper as no new data were created or analyzed in this study.

This work was funded by National Key R&D Program of China (2024YFB3814000) and R&D Program of Zhejiang (2024C01242(SD2)).

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.