This document summarizes a study that uses SPICE simulations to analyze the behavior of photovoltaic systems under various ground fault conditions by changing the ratings of ground fault protection devices (GFPDs). The simulations show that decreasing the GFPD ratings improves the detection of high impedance, low current ground faults without increasing nuisance trips. Specifically:
1) A SPICE model of a PV array was created using a single diode module model to simulate the behavior of the array under different fault conditions.
2) Simulations were run for faults at different module positions in arrays of varying sizes. The results show the GFPD current is sub-linear with respect to array size.
3) For large arrays, decreasing
This document summarizes a study that uses SPICE simulations to analyze the behavior of photovoltaic systems under various ground fault conditions by changing the ratings of ground fault protection devices (GFPDs). The simulations show that decreasing the GFPD ratings improves the detection of high impedance, low current ground faults without increasing nuisance trips. Specifically:
1) A SPICE model of a PV array was created using a single diode module model to simulate the behavior of the array under different fault conditions.
2) Simulations were run for faults at different module positions in arrays of varying sizes. The results show the GFPD current is sub-linear with respect to array size.
3) For large arrays, decreasing
This document summarizes a study that uses SPICE simulations to analyze the behavior of photovoltaic systems under various ground fault conditions by changing the ratings of ground fault protection devices (GFPDs). The simulations show that decreasing the GFPD ratings improves the detection of high impedance, low current ground faults without increasing nuisance trips. Specifically:
1) A SPICE model of a PV array was created using a single diode module model to simulate the behavior of the array under different fault conditions.
2) Simulations were run for faults at different module positions in arrays of varying sizes. The results show the GFPD current is sub-linear with respect to array size.
3) For large arrays, decreasing
Sandia National Laboratories, Albuquerque, NM 87185, USA
fault conditions [3]. A common method of circuit simulation
Abstract — In order to better understand the impact of is the use of the Simulation Program with Integrated Circuit changing the fuse ratings on ground fault protection devices Emphasis (SPICE). SPICE, originally developed at the (GFPDs) of photovoltaic (PV) arrays with regard to detection University of California, Berkeley Electronics Research window and nuisance tripping, a range of fault scenarios, including array size, leakage current, fault position, and fault Laboratory in 1973 [4], is a general-purpose, open source, impedance were analyzed using a simulated program with analog circuit simulator used to predict circuit behavior. In integrated circuit emphasis (SPICE). The simulations show that, this work, the program MacSPICE, a derivative of SPICE3f4, by decreasing GFPD ratings, there is an improvement in the is used to analyze the behavior of PV systems in various array number of faults that can be detected without increasing the configurations and ground-fault conditions [5]. number of nuisance trips and, therefore, the GFPD ratings The SPICE model of the PV array (Figure 1) is mandated by UL1741 should be decreased in the future. accomplished through the construction of a single module Index Terms — SPICE, ground fault, PV, GFPD, GFDI using a one-diode model. This module consists of an ideal current source at the module short-circuit current (Isc) in I. INTRODUCTION parallel with a diode and shunt resistance (Rsh) and in series with a series resistance (Rs). In order to increase the Voc of the The mechanical, electrical, or chemical degradation of PV module above the voltage drop of a regular diode (~0.6 V), the systems can lead to a number of fault scenarios, including ideality constant of the diode is increased [6]. series faults, ground faults, intra-string faults, and string-to- string faults. The behavior of the array and the fault current/voltage can vary widely dependent on the type, location, and impedance of the fault path, making accurate detection without false-positives (nuisance tripping) challenging. In order to mitigate ground fault effects, a GFPD—often called the Ground Fault Detector/Interrupter (GFDI)—is used. These devices typically consist of a simple 10x38 mm “midget” style fuse. Figure 1: (left) One diode model for single PV module. (right) IV While there are a number of PV configurations and curve of single PV module topologies in the United States, the most common is the grounded, isolated system. Unfortunately this design is For the purposes of this work, the one-diode model is known to have problems with undetectable ground faults in constructed to approximate a nearly perfect photovoltaic some cases [1, 2] where the GFPD is not capable of de- module. The current source is set to supply 2.5 A at short energizing the fault, resulting in multiple fires. Industry circuit, the diode has an ideality factor of N=80, the shunt leaders have suggested lowering the GFPD (i.e., fuse) ratings resistance is set 11020 and the series resistance is set to 10 would improve the chances of detecting high impedance, low m. This module gives an IV curve with Isc of 2.5 A, Voc of current ground faults. However, many system designers and 56 V, and Pmp of 118 W. The max power point has a current operators are hesitant due to the belief that this will increase of 2.4 A and a voltage of 49.2 V. nuisance-tripping events. In this work we conduct SPICE The PV array model is comprised of a number of strings simulations in order to optimize the GFPD fuse rating (and wired in parallel. Each string is composed of 7 modules in thus increasing the fault detection window) without series (Figure 2). Each module is connected to a bypass diode exacerbating nuisance tripping. (Isat=4.710-12 A, N=1). When multiple modules are placed in series to form a string, the frames of all the modules are II. PV MODEL connected to earth via equipment grounding conductors [7]. This results in leakage current from the module to ground. In PV modules are non-linear circuits. Their behavior is this work, to model leakage current, the positive terminal of difficult to describe analytically without transcendental each module in a string has a leakage resistor connected to equations. However, the use of computer circuit simulations ground. can describe the behavior of a PV system for a wide variety of While there is current leakage from a number of places in this abstract only details related to GFPD sizing will be the array, the primary source of leakage in a PV array is in the discussed. modules. Module leakage current is defined as the current between the module, biased at some high voltage, and the III. RESULTS AND DISCUSSION grounded frame. Leakage current measurements for crystalline Silicon modules are described by IEC 61215 A set of simulations was carried out for a 25 fault at the (section 10.15.3) [8] and UL1703 (section 21.5) [9]. In order positive terminals of modules 1 through 7, denoted as 1+ to simulate the leakage current of a real array, the positive through 7+ (Figure 3). The GFPD current is sub-linear with terminal of each module in a string has a leakage resistor respect to array size and has a near linear relationship with (Rleak) connected to ground. Rleak is taken to be a value of 5 array size for small array sizes and a nearly independent M, giving a 100 A leakage current for a module biased at relationship for large array sizes. This is due to the inverter 500 V (UL maximum leakage current for a 1.5 m 2 module is Rmp decrease as array size increase. As strings are added in 11.1 A at 500 V). parallel, the PV current is increased without a corresponding increase in output voltage. Since R mp has an inverse dependence on system current, Rmp decreases rapidly for smaller array sizes while remaining fairly constant for large array sizes.
Figure 2: Schematic for a two-string array of modules that have a
leakage current, Rleak. The red lines denotes the current path through the GFPD for a fault to ground. Figure 3: GFPD current vs. array size for 25 fault at various module positions. The GFPD current is clearly sub-linear with In each simulation, the PV array is constructed with respect to array size. For large array sizes, the load resistance (Rmp) multiple modules connected to a central inverter. A grounded, is small enough compared to the fault resistance to divert a isolated, central inverter is modeled because this topology is significant amount of current from the GFPD through the load. susceptible to faults that are undetected by the ground fault fuse. A resistor is taken as a basic approximation of the real For small arrays, the value of R mp is large compared to the input impedance of an inverter and is set at the resistance fault resistance, so that the current path through the required to generate maximum PV power, R mp, of the inverter/load is much more resistive than the current path unfaulted array. through the fault/GFPD. As more strings are added, most of Although physical inverters are a complex system the added current flows through the fault path. containing transistors, capacitors, and switching controls [10], For large array sizes, the value of Rmp has decreased so that only the real component of the inverter impedance is modeled the resistance of the inverter current path is smaller than the here since, if voltage ripple and max power tracking [11] are resistance of fault/GFPD path. As more strings are added, ignored, the PV system can be approximate by a steady state nearly all of the added current flows through the inverter and DC system with a constant resistance. This means that the the GFPD current is nearly constant with regards to array size. imaginary components of the inverter impedance have no UL1741 [12] defines the maximum GFPD detection setting effect on the steady state solution before and after the fault as a function of array size. The standard lists a maximum occurrence. During changes in state of the array, transients rating of 1 A for array sizes 0-25 kW, 2 A for 25- 50 kW, 3 A due to energy storage elements are not long lasting enough to for 50-100 kW, 4 A for 100-250 kW, and 5 A for arrays trip GFPDs. greater than 250 kW. The simulation data shown in Figure 3 The SPICE simulations carried out in this work have shows GFPD currents for certain fault conditions which would focused on various aspects of GFPD sizing related to the PV not meet the maximum GFPD detection settings. array, including array size, module leakage, GFPD resistance, For example, for arrays in the 100-250 kW range, the GFPD and fault location. However, due to space considerations in rating is allowed to be upwards of 4 A. SPICE simulations of ground faults have shown that, in this case, the fault current for a ground fault at the positive terminal of either of the first ACKNOWLEDGEMENTS two modules (1+ or 2+) is below this threshold of protection. Most disturbing is the 1+ fault cause, which provides less This work was funded by the DOE Office of Energy than 1 A of GFPD current for all array sizes. This GFPD Efficiency and Renewable Energy. Sandia National threshold is undetectable by the UL1741 standard for GFPD Laboratories is a multi-program laboratory managed and rating for any array size. operated by Sandia Corporation, a wholly owned subsidiary of Figure 4 shows a close-up view of the 1+ fault shown in Lockheed Martin Corporation, for the U.S. Department of Figure 3. The GFPD current of a 1+ fault (red trace) shows a Energy's National Nuclear Security Administration under very small dependence on array size for large arrays. The contract DE-AC04-94AL85000. entirety of this slope is due to module leakage current (blue trace). It is apparent that the UL1741 standards would not detect such a fault for any array size. In order to fully protect REFERENCES a PV system from a 25 fault, it would be required to have a 0.25 A fuse for 1- and 2-string arrays (IGFPD=0.30 and 0.45 A). [1] B. Brooks. (2012, Mar) Bakersfield Report SolarPro. 62- Larger array sizes could utilize a 0.5 A fuse if there were 70. extremely high module leakage currents, e.g., 1 mA (~120 [2] A. Rosenthal, G. Ball, B. Brooks, and J. C. Wiles, "The SolarABCs Ground-Fault Research Project - Examination times larger than the IEC 61215 requirement). However, a of Inverter Ground-Fault Detection Blind Spot and 0.25 A fuse would almost certainly not experience nuisance Recommendations for Mitigation," In Press. tripping from leakage current alone (shown by the 100 A [3] Y. Zhao, B. Lehman, J. de Palma, J. Mosesian, and R. blue trace) even for large array sizes. Lyons, "Challenges to overcurrent protection devices under line-line faults in solar photovoltaic arrays," Audio, Transactions of the IRE Professional Group on, pp. 20-27, Sep 01 2011. [4] L. W. Nagel and D. O. Pederson, "SPICE (Simulation Program with Integrated Circuit Emphasis)." vol. ERL- M382, ed: University of California, Berkeley, 1973, p. 62. [5] L. Castaner and S. Silvestre, Modelling Photovoltaic Systems using PSPICE. Chichester, West Sussex, England: John Wiley and Sons Ltd, 2002. [6] A. Devasia and S. K. Kurinec, "Teaching solar cell I-V characteristics using SPICE," American Journal of Physics, vol. 79 (12), p. 1232, 2011. [7] J. C. Wiles, "Photovoltaic System Grounding," ed, 2012, pp. 1-29. [8] International Electrotechnical Commission, "Crystalline Silicon Terrestrial Photovoltaic (PV) Modules-Design Qualification and Type Approval," 2nd ed. Geneva, p. 102. Figure 4: Close-up view of the 2+ and 1+ fault positions. The 1+ [9] Underwriters Laboratories, "Flat-Plate Photovoltaic fault position has a sub-linear dependence on array size with a smal Modules and Panels," 3rd ed. Northbrook, IL, 2008, pp. 1- slope at large array sizes (due entirely to module leakage current). 82. For any array size, a 1 A fuse (the lowest fuse rating mandated by [10] J. Worden and M. Zuercher-Martinson. (2009, May) How UL1741) would be insufficient to catch such a fault. Inverters Work. SolarPro. 68-85. Available: IV. CONCLUSION http://xiazai.inktronics.com.cn/Uploadfile/DownFile/ 10-08-13/100813131907_hjkg.pdf [11] J. Flicker, R. Kaplar, M. Marinella, and J. Granata, "PV Undetected fault conditions in PV arrays present dangers to Inverter Performance and Reliability: What is the Role of normal PV operations because they have been the cause of the Bus Capacitor?," Photovoltaic Specialists Conference fires in the past. This is especially problematic for roof- (PVSC), Austin, TX, 2012, p. 6. mounted systems. It is vital to mitigate fire hazards by [12] Underwriters' Laboratory, "Inverters, Converters, increasing the fault detection window and decreasing the Controllers, and Interconnection System Equipment for changes of an undetected fault condition. However, care must Use with Distributed Energy Resources," UL 1741 ed. be taken to increase detection without a significant increase in Northbrook, IL, 2001. nuisance tripping due to transients or module leakage. In this work, SPICE simulations of ground fault conditions of arrays of various sizes have shown that even modestly resistive ground faults can be undetected if the UL1741 maximum allowable GFPD ratings are used. The results strongly suggest that GFPD ratings can be lowered below 1 A, even for large array sizes, without inducing nuisance tripping due to module leakage.