1 Lightning Protection for Power Systems: A Primer for Undergraduate Students Michael J. Bloem, Member, IEEE Abstract—This paper presents an overview the phenomenon of lightning and its effect on power systems for both direct and indirect strokes. It also reviews the behavior, characteristics, and physical implementation of surge arrestors, which are commonly used to protect power systems from lightning. Modeling and simulation of the use of surge arrestors was performed using MathCad and PSCAD software. Finally, more practical tools for choosing and implementing an arrestor were reviewed, and an actual arrestor was chosen for a typical situation. Index Terms—industrial power systems, lightning, surge protection provides the impetus for a “stepped ladder” of current to begin flowing down to earth. Essentially, the collection of negative charge makes an attempt to discharge to ground when encouraged by a pocket of positive charge within the cloud. The stepped ladder typically flows in steps of tens of meters, and has pulse currents of more than 1 kA. The potential difference between the stepped ladder and the ground can reach 100 MV. Eventually, the stepped ladder reaches ground, usually by attaching to an upward streamer that rises up from the ground. Once the stepped ladder does connect to the ground, the negative charge that was lowered via the stepped leader is discharged. Fig. 1 depicts this process. I. INTRODUCTION L strokes are more than a nuisance to power systems engineers. Lightning accounts for around a third of all power outages in the United States, leading to many millions of dollars in costs. In order to protect power systems from lightning strokes, the strokes themselves and how they impact power systems must be analyzed. The resulting understanding allows a power system engineer to pick the right form or protection for the power system in question, which will more often than not be a surge arrestor. Once the operation of surge arrestors is understood and the appropriate surge arrestor characteristics have been determined for a certain application, an actual arrestor must be located and purchased. This paper summarizes research and analysis that covers all of these various topics. IGHTNING II. THE PHENOMENON OF LIGHTNING Lightning in generated as a result of current being separated by wind up and down drafts. These drafts leave negative charges at the bottom of a thundercloud and positive charges at the top. This collection of negative charge at the bottom of the cloud is the ultimate cause of cloud-to-ground lightning. A cloud-to-ground stroke begins when there is a preliminary breakdown at the bottom of the cloud. This breakdown involves a small pocket of positive charge that M. J. Bloem is a student at the Electrical Engineering Department, Calvin College, 3201 Burton St SE, Grand Rapids, MI 49546 (e-mail: mbloem63@calvin.edu). Fig. 1. Formation of a stepped leader that starts a lightning strike. Note how the preliminary breakdown is caused by a pocket of positive charge within the cloud [2]. At this point, a breakdown current pulse referred to as a return stroke forms. The return stroke travels back up the channel created by the breakdown current pulse. This current typically flows at a third of the speed of light and has a current of 30 kA. Obviously, a current of this magnitude generates magnetic and electric fields, which can last for several milliseconds and cause damage to power system equipment. A sequence of diagrams in Fig. 2 show the sequence of events involved in a lightning strike, from stepped leader to return stroke. Fig. 2. A series of diagrams depicting a stepped leader coming down from a cloud and how it connects with an upward leader coming up from the ground. In the later diagrams, the return stroke is shown [2]. 2 When processes referred to as “J” and “K” processes occur within the thundercloud, more charge can become available at the cloud side of the channel created by the stepped leader. This charge can lead to 3-5 subsequent strokes, which often follow the path of the first stroke. These strokes have lower current levels, but are faster (in terms of speed to peak current). The interval between these subsequent strokes is about 60 milliseconds [2]. III. THE EFFECT OF LIGHTNING ON POWER SYSTEMS Clearly, an electrical phenomenon as powerful as a lightning strike can have a devastating impact on power systems. In the United States, 30% of all power outages are lightning related. The total cost related to these outages is greater than 1 billion dollars. There are many factors, however, that influence the likelihood that a power system will be struck by lightning. Electrical power engineers have developed a few parameters that indicate the how much danger a certain power system faces from lightning strikes. The first of these parameters is ground flash density (GFD). The GFD is the number of lightning flashes striking the ground per unit area per year. This is a value is calculated based on measurements taken over a large number of years. It varies quite widely with climate. For example, the GFD is as much as 14 flashes/km/year in parts of Florida and less than 0.1 flashes/km/year in parts of California [2]. A related parameter that is more specific and useful is the incidence of lightning to power lines. This parameter is defined as the number of flashes striking the line per 100 km per year. It is comparable to the fault index due to direct lightning hits for distribution lines, and is an indicator of the exposure of the line to direct strikes. The equation used for the determination of the incidence of lightning to power lines is shown below. 28 h 0.6 b N GFD 10 (1) Where N is the incidence of lighting to power lines, GFD is the ground flash density, h is the pole or tower height, and b is the structure width. A third parameter that tells just how much energy is dissipated once a power system is struck by lightning is the action integral. More specifically, the integral tells the amount of energy that would be dissipated in a 1 Ω resistor if the current from the lightning strike were to flow through it. The median value for the action integral is 650,000 A2s. This parameter is gives power system engineers an idea of how much damage will be done to systems (and surge arrestors) that are actually struck by lightning. A. Direct Lightning Strokes Lighting strokes have different impacts on power systems depending on if they are direct or indirect. A stroke is direct when it hits the phase conductor, the shield wire, or the tower. A few terms and parameters should be introduced relating to the impact of direct lightning strokes on power systems. “Backflash” refers to the insulating string at a tower flashing over to the tower or shield line. “Shielding failure” occurs when the insulating string flashes over by a strike to the phase conductor. The basic impulse insulation level (BIL) is an important parameter that indicates the size of a voltage impulse that a particular insulation implementation can handle (without flashing over). There are two types of BIL: the statistical BIL and the conventional BIL. The statistical BIL is the crest value of a standard lightning impulse voltage that the insulation will withstand with a probability of 90% under specified conditions. The conventional BIL, on the other hand, is the crest value of a standard lightning impulse voltage that the insulation will withstand for a specific number of applications under specified conditions. So the statistical BIL is relates to a one-time withstanding of a voltage, while the conventional BIL is a long-term parameter relating to the same thing. 1) Direct Strokes to Unshielded Lines Direct lightning strikes have different impacts depending on if they hit the phase inductor or the tower. When a stroke hits a phase inductor, the return-stroke splits into to halves, each traveling one direction down the line. This produces traveling voltages of magnitude V Z0 I 2 (2) Where I is the return stroke current and Z0 is the surge impedance of the line. The surge impedance is in turn calculated using equation (3). L Z0 C 1/ 2 (3) C and L are the series inductance and capacitance to ground per meter length of wire. This voltage is attenuated as it travels down the line. If the voltage exceeds the BIL, then a flashover might occur and result in a power outage. The critical return stroke current (the return stroke current that will cause an inductor flashover) can be calculated. IC 2 BIL Z0 (4) Before the impact of a direct lightning stroke to a tower (as opposed to a phase inductor) can be analyzed, one must decide how to model a tower. Short towers can simply be ignored, but not longer towers. Typically, the tower is modeled as a vertical transmission line with surge impedance Zt, where the voltage and current waves travel with a velocity of vt. The tower is terminated at the lower end by the towerfooting resistance, Rtf, and at the upper end by the lightning channel, with another surge impedance of Zch. Using this model, when a tower is struck, voltage and current waves will repeatedly travel up and down the tower, reflecting back at each end. This leads to a voltage at either end of the tower cross-arm. The insulator voltage will be equal to the cross-arm voltage if the power-frequency voltage 3 is ignored. In the case of a direct lightning stroke to a tower, this is the voltage that must remain below the BIL in order to avoid a flashover. 2) Direct Strokes to Shielded Lines Shielding lines are conductors strung above phase conductors of overhead power lines. These are directly attached to towers so that return-stroke currents are safely led to ground through the tower-footing resistance. Moreover, the critical current for a shielded line is higher than that for an unshielded line because the presence of the grounded shield wire reduces the effective surge impedance. Yet another advantage of shielding wires is that the return stroke current is divided up into three parts (tower and each direction on the shielding wire), so a lower voltage will be developed across the tower and each part of the shielded line. Ultimately, this means that a higher voltage can be handled by the line. Finally, one last reason that shielding wires reduce the voltage that must be sustained by the insulator for a given return stroke is that electromagnetic coupling between the shield wire and the phase conductor induces a voltage on the phase conductor, thus lowering the voltage difference across the insulator. Fig. 3 a and b show the difference in magnitude of insulator voltage experienced by (a) unshielded versus (b) shielded lines. B. Indirect Lightning Strokes Indirect lightning strokes do not actually physically connect with the phase inductors, shield line, or tower, so they induce overvoltages differently than direct strokes. More specifically, there are four components to the voltage induced by an indirect stroke in a phase inductor. The first component is a result of the charged cloud above the line inducing bound charges on the line while the line itself is held electrostatically at ground due to leakage over the insulators and the neutrals of connected transformers. When the cloud is discharged, these bound charges are released, giving rise to voltages and currents. The second component of the voltage induced on a line by an indirect lightning stroke is induced by the charges lowered by the stepped ladder. When the charges in the line bound by the charges lowered by the stepped ladder are released as the return-stroke neutralizes the stepped ladder, voltages and currents in the line result. Electrostatic fields in the vicinity of the line are induced by residual charges in the return stroke. This leads to the third component of the voltage induced on a line by an indirect lightning stroke. Finally, the last component is magnetically induced due to the rate of change of current in the return stroke. Of these four components, only the last two are significant and the first two can be neglected safely. The inducing voltage is an important parameter in the calculation of overvoltages resulting from indirect strokes. This is the voltage at a field point in space with the same coordinates as a corresponding point on the line conductor, but without the presence of the line conductor. It is calculated by first finding the total electric field created by the charge and the current in the lightning stroke at any point in space. Ei E ei E mi (a) A t (5) In equation 5, Φ is the inducing scalar potential created by the residual charge at the upper part of the return stroke (the third component listed above). A is the inducing vector potential created by the upward moving return-stroke current (the fourth component listed above). In order to find the inducing voltage Vi, the line integral of Ei must be taken, as dictated by the potential gradient relationship between potential and electric field intensity. hp Vi Ei dz Vei Vmi (6) 0 (b) Fig. 3. Difference in magnitude of insulator voltage experienced by (a) unshielded and (b) shielded lines [3]. Of more interest than the inducing voltage is the induced voltage. This is the voltage that is actually induced in the conductor as a result of the inducing voltage. The two voltages will be different because current can be conducted when the conductor is present. Unfortunately, the calculation of the induced voltage from the inducing voltage is complex mathematically and will not be pursued extensively in this paper. Some possible tools to use in this calculation include Dahamel’s integral and Green’s function. 4 The calculation of outage rates caused by nearby lightning strokes is more straightforward. The number of flashovers (nfo) for a particular region is a function of the magnitude of the return stroke current (ΔIp), the front-time step of the line (Δtf), the probability that a strike big enough and with enough of a front-time step will hit in the right spot (p(Ip,tf)), the ground flash density of the region (ng), and the attractive area (A) [2]. nfo p( I p , t f ) I p t f n g A (7) IV. THE USE OF SURGE ARRESTORS TO PROTECT POWER SYSTEMS AGAINST LIGHTNING A. Surge Arrestors Surge arrestors are place in parallel with a power system object to be protected and in close proximity with it. The idea is to make sure that the voltage across the arrestor never exceeds what the protected object can handle [7]. Surge arrestors are ideal for this task because they have a non-linear relationship between voltage and current. When the voltage across the component to be protected is relatively low, the surge arrestor has a high impedance and therefore does not allow much current to flow through it. This keeps power consumption low under normal operating conditions. When a voltage surge occurs, however, the surge arrestor allows a large amount of current to flow, limiting the voltage across itself and the protected component [1]. 1) Behavior The most important behavior of a surge arrestor is its non-linear relationship between voltage and current, as mentioned above. The high resistance in normal operation minimizes steady-state losses, while the lower resistance level at higher voltages provides an alternate discharge path for surge currents that keeps system voltages at safe levels [1]. The V-I characteristic of a surge arrestor can be classified into three main regions. The first region is the leakage region. In this region the arrestor draws a small current (typically less than 0.1 mA/cm2). During steady-state operation the arrestor should be operating in the leakage region. As voltage levels increase, the second region is reached. This region is referred to as the transition region, where the nonlinear behavior of the arrestor starts. Technically, the “transition voltage” that separates the leakage region from the transition region starts wherever the current density becomes 1 mA/cm2. Finally, the highly nonlinear region is the reached as voltages continue to increase. This region is characterized by large incremental current changes for small incremental voltage changes. The highly nonlinear region is also characterized by a high thermal coefficient, which means that thermal runaway is a danger in this region [1]. Another important characteristic of the behavior of a surge arrestor is the amount of energy it can dissipate. An arrestor must be able to dissipate the energy of the surge without damaging itself [1]. 2) Physical Implementation The most basic surge arrestors are implemented with a combination of gaps and silicon carbide resistors. Silicon carbide is the material with the nonlinear relationship between voltage and current in the arrestor. Gaps are used to prevent any current from flowing while the device is in its steady-state mode of operation. When a voltage surge does occur, the gap sparks over and the arrestor is able to control the voltage across the protected component [1]. Almost all modern surge arrestors are implemented with stacks of metal-oxide varistors (MOV) as the nonlinear element. Fig. 4 shows the superior nonlinear characteristics of a MOV arrestor as compared with a SiO arrestor. MOV arrestors quickly allow more and more current to flow as the voltage level increases, meaning that a MOV arrestor can be chosen to allow virtually no current to flow during steady state operation and very large amounts to flow during overvoltages. The performance is such that a gap in series with the material at all because so little current flows at lower voltages. Fig. 4. Nonlinear characteristics of an MOV arrestor and an SiO arrestor [7]. The metal oxide most commonly used in MOVs is zinc oxide, which is used in 90% of all MOVs. The disks-shaped blocks of ZnO are made by a process involving grinding, mixing, pressing, and sintering. The resulting material is dense and fine. The nonlinear properties actually result from the boundary layers of the crystals. The grain size is dependent on the sintering process, and grain size determines how many boundaries there will be. In this way the production process can be altered to produces different nonlinear characteristics. The dialectic properties of the material determine the number of disks that will be required. Often several columns will be applied in parallel to the protected component. This is helpful when more energy needs to be dissipated than can be handled by a single column. Fig. 5 shows in detail the various components of one implementation of a surge arrestor. 5 Fig. 5. Schematic diagram of valve type arrestor [7]. An arrestor involves more than just a metal oxide nonlinear resistor. Other components include gaps, magnetic coils, and shunting resistors. The two gap units flash over when the voltage level increases rapidly as a result of a lightning strike. The breakdown of the gaps is actually precisely controlled with the use of pre-ionizing tips, which maintain a higher-thannormal number of electrons in the gap, ready to ignite a breakdown. The bypass gap in parallel with the magnetic coil flashes over when a large surge of current flows down through the arrestor and around the coil, which does not conduct transient currents well. At this point the voltage across the arrestor is basically controlled by the nonlinear thyrite resistor. Once the surge has passed, the impedance of the magnetic coil becomes much lower as more DC-like conditions resume. This causes the gap in parallel with the coil to diminish and die. Flux develops as a result of the current flowing through the coil, which pushes the arcs in the nearby gap units into quenching chambers. These quenching chambers elongate and cool the arc as it is eliminated. The arrestor returns to its quiescent state. An alternate implementation of a gapped surge arrestor is depicted in Fig. 6. In this case, the voltage distribution between the gap and the MOV when in quiescent mode is determined by the capacitance across the gaps (C1) and the inherent capacitance of the MOV (C3). This distribution impacts the power consumption during steady-state operation. When a surge is applied to the arrestor, C2 acts as a short across gap 2. This forces gap 1 to short as it must then handle a large voltage on its own. Once it has shorted, gap 2 sparks over and the MOV takes control of the voltage across the arrestor [7]. Fig. 6. Circuit diagram of a gapped surge arrestor [7]. 3) Potential issues There are two main aspects of surge arrestor usage that could be problematic for engineers. The first such area is energy dissipation. If an arrestor is asked to dissipate too much energy, it will not return to its high-resistance state when the overvoltage is complete. This results in an unwanted short circuit, which could devastate a system. Thermal runaway is the second area of concern for engineers. Thermal runaway is a positive-feedback phenomenon where high current levels heat up the nonlinear resistor, which leads the arrestor to conduct even more current. Ultimately, the level of current may require the arrestor to dissipate more energy than it is able to, damaging the arrestor. In order to prevent these aspects from becoming problematic, engineers must ensure that the current and energy dissipation capabilities of the arrestors they choose to use will fulfill the requirements of the system they are designing for [1]. B. Mathematical Description of Surge Arrestors The V-I characteristic of a surge arrestor is given by equation 8. V I (V ) p sgn( V ) (8) V0 Where I is the discharge current of the arrestor, V is the discharge voltage of the arrestor, V0 is the transition voltage of the arrestor, α is a parameter that describes the sharpness of the nonlinear region of the arrestor (typically between 10 and 50), and p is a parameter than can be adjusted to allow for leakage current under normal voltage conditions. An example application of this equation will be given. Suppose we have an arrestor with the following characteristics: Disk diameter, D = 50 cm2 Disk thickness, h = 2.1 cm2 DC voltage at 1 mA/cm2, E = 1640 V/cm 6 DC leakage current density at 80% of DC voltage at 1 mA/cm2, J = 2 μA/cm2 Voltage ratio at 300 A/cm2 to 1 mA/cm2, V300_1 = 1.70 System operating voltage, Vop = 138 kV Maximum sustainable steady state current density, Imax = 1 mA/cm2. Note that the characteristics are given in this case with respect to Imax, which is dependent on the material used in the arrestor. The transition voltage of the arrestor, V0, is calculated as the product of the thickness of the disk and the DC voltage at 1 mA/cm2 (see equation 9). V0 h E Energy dissipation in surge arrestors is also important to study because excessive levels of energy dissipation can lead to overvoltages if of long enough duration. Even relatively slow transients can become problematic. To illustrate how a system with a surge arrestor responds to a transient, the surge arrestor column analyzed above will be applied to a simple single-phase system undergoing a no-load bus energization. The system in question is shown in Fig. 8. The source voltage is applied at t = 0. (9) The nonlinear sharpness parameter α is a function of V300_1 and is calculated using equation 10. 300 ln 0.001 ln V300 _ 1 (10) Fig. 8. Simple single-phase system undergoing a no-load bus energization. The source voltage Vs is applied at t = 0. The value for p, the parameter that accounts for leakage current under normal operating conditions, depends on the disk diameter and the current density, as shown in equation 11. J D p 0.8 (11) Enough surge arrestor disks must be placed in series in order to limit the steady state current density to Imax. The number required (n) can be calculated if V0 and V300_1 are known (see equation 12). 2 Vop 1.5 n ceiling 3 V0 (12) Once the required number of disks is known, the value of V0, the transition voltage, must be multiplied by this number in order to get the new individual containers. Using these equations, the V-I characteristic can be plotted. For this example, the characteristic is shown in Fig. 7. Fig. 7. V-I characteristic of the example surge arrestor column. Note how the current is nearly zero for a wide range of voltages, but when the voltage level gets high enough, current is allowed to flow so that the voltage does not become too high. The parameters for this system will be Resistance, R = 3.5 Ω Inductance, L = 35 mH Capacitance, C = 70 µF Source voltage, Vs (t ) 2 Vop cos(t ) kV 3 ω = 377 radian/sec φ = 10 degrees In order to solve this system, the derivatives of the current and the voltage are determined. Vs V R I L (13) L I I (V ) dV L (14) C dI As the energy dissipation in this system will be determined numerically, the time step and duration must be set. The time step (dt) will be 0.1 ms and the duration (T) will be 100 ms. N will denote the required number of solution points and is simply T divided by dt. The auxiliary constants used in the numerical integration will be 1.5 and 0.5 (h1 and h0). Next, the initial conditions must be set. Inductor current, IL0 = IL1 = 0 A Arrestor and capacitor voltage, V0 = V1 = 0 V Arrestor energy, E0 = E1 = 0 J Finally, the numerical integration is performed and the arrestor current, voltage, and energy dissipation are determined. I Li1 I Li h1 dI i h2 dI i 1 (15) Vi 1 Vi h1 dVi h2 dVi 1 (16) Ei 1 Vi I (Vi ) dt (17) The arrestor voltage is shown in Fig. 9. 7 C. Simulation of Surge Arrestor Protection in PSCAD To get an idea of how a surge arrestor actually works when applied to a power system during a lightning strike, a simulation was performed in PSCAD. PSCAD is a software package that was developed at the University of Manitoba to simulate power systems. The PSCAD sheet used is shown in Fig. 12. 4 10 5 2 10 5 Vi 0 2 10 5 4 10 5 0 0.02 0.04 0.06 0.08 0.1 ti voltage Fig. 9. Arrestor voltage during no-load bus energization. Fig. 3.3.4 Voltage across arrester The arrestor current is shown in Fig. 10. 50 0 50 I Vi 100 Fig. 12. PSCAD model of a power system. A lightning strike is applied to the system, and the response of the system is examined. A surge arrestor protects the system from overvoltages or flashover. 150 200 0 0.02 0.04 0.06 0.08 0.1 ti current Fig. 10. Arrestor discharge current during no-load bus energization. Note the transient current that passes throughdischarge the arrestor ascurrent the voltage is Fig. 3.3.5 Arrester controlled during the energization. Lastly, the arrestor energy dissipation is shown in Fig. 11. 6000 3 4.86410 4000 There are a number of aspects of this system that are worth pointing out. The two cables on the left side of the diagram represent a transmission line. One cable is a service cable and the other is a ground cable. The two parallel lines at the top of the diagram represent the high voltage and neutral line in the system. The three resistors slightly right of center in the diagram represent the transmission line. In this case the transmission line is represented simply as a surge impedance because for the time frame we are concerned with here, the line will behave according to its surge impedance. In other words, the simulation time is less than the travel time for the line, so the line is just a surge impedance in this simulation. The subsystem at the far right of the diagram represents the lightning strike that is applied to the power system. The lightning surge is represented by equation 18. Ei I 50 e5010 0 t e1.210 6 t (18) This is a standard form used to describe a lightning surge. The item in the upper middle of the diagram is a steepfront surge arrestor. It is actually a subsystem with separate components, as seen in Fig. 13. 2000 0 6 0 0.02 4 110 0.04 0.06 ti 0.08 0.1 0.1 energy Fig. 11. Arrestor energy dissipation during no-load bus energization. Fig. 11 shows that the arrestor must be able to dissipate 55 kJ of energy during the transient. This energy is dissipated as the arrestor allows current to flow through it in order to limit the voltage applied to the bus during the initialization [1]. 8 voltage between the high and neutral lines. So much current is flowing through the arrestor at higher voltages that the voltage should not remain at more than 2.5 or 3 times the operating voltage for any significant length of time. The results of the simulation are shown in the series of graphs shown in Fig. 15 and Fig. 16. Fig. 15 shows the current, in kA, flowing through the arrestor over time. Note how the current increases in response to the lightning strike transient, thereby limiting the voltage between the high and neutral lines. Fig. 13. PSCAD model of a steep-front surge arrestor. The arrestor is actually constructed of two arrestors in parallel, which allows the arrestor to handle particularly fast transients. The most important characteristic of this arrestor to note is that it actually combines two arrestors in parallel. This allows the arrestor as a whole to handle more current and therefore limit voltages even during particularly fast and powerful transients. The sharing of current allows each arrestor to dissipate less power, making thermal runaway and damage to the arrestors less likely. The V-I characteristic of the surge arrestors was set using a dialog box as shown in Fig. 14. Fig. 15. Current across the surge arrestor component over time in the PSCAD simulation. The current increases rapidly in response to the lightning strike, thereby limiting the voltage between the high and neutral lines. Fig. 16 shows the voltage between the high and neutral lines of this system. Initially, this voltage is fairly large (70 kV) relative to the operating voltage (5 kV). However, the steep-front surge arrestor quickly kicks in and limits the voltage to around 12 kV by the time 4 µsec have elapsed. This is about 2.4 times the operating voltage, and is a reasonably safe level of operation. Fig. 14. PSCAD dialog box to set up the V-I characteristic of the surge arrestors. The column on the left shows current levels at various multiples of the operating voltage, which is in the column on the right. Note that at low multiples of the operating voltage (which is 5 kV in this simulation), the current flow is very small. However, the current flow becomes significant at around 1.75 times the operating voltage. By allowing current to flow through itself at higher voltages, the surge arrestor controls the Fig. 16. Voltage between the high and neutral lines over time in the PSCAD simulation. Note how the voltage is limited to 12 kV (2.4 time the operating voltage) after just 4 µsec. The effect of the surge arrestor in this situation is best understood by considering what the result would be if the surge arrestor was not present. When the simulation is run with the surge arrestor removed the voltage between the high 9 and neutral lines reached excessive levels. As can be seen in Fig. 17, the voltage between these lines approaches 8 MV, a dangerously high level that would certainly lead to damage to the power system [6]. Fig. 17. Voltage between the high and neutral lines over time in a PSCAD simulation without a surge arrestor present. Note how the voltage reaches extremely high levels (8 MV) relative to the operating voltage (5 kV). D. Practical Advice for Arrestor Selection and Implementation The proper approach when selecting and implementing a surge arrestor is to first determine the minimum arrestor rating that will not lead to damage to the arrestor. One the arrestor has been selected, then the insulation level that is required to result in an acceptable probability of flashover or risk of failure must be defined. Essentially, the current and voltage flowing through the arrestor are controlled by the selection of the arrestor, but the insulation level determines if the system will be able to handle the voltage level that the arrestor limits to. For example, an arrestor might hold the voltage level relatively low, but if the insulation is very poor, flashover could still occur. There are four main factors of an application to which an arrestor is to be applied to consider when selecting an arrestor. The first factor is the maximum fundamental frequency continuous operating voltage (MCOV) applied to the arrestor. The MCOV rating of the arrestor must be at least equal to and should usually exceed the highest continuous system voltage by some small margin. The temporary fundamental frequency overvoltage (TOV) that the arrestor will be exposed to is the next quantity to consider. During fault conditions or when a lightning strike has occurred, the line-to-ground voltages on unfaulted phases can rise significantly. This value gives the level of overvoltage that can be expected for a short period of time, and the chosen arrestor must be able to withstand this level of voltage. The third factor to consider is the energy that the arrestor must absorb during a lightning strike. As mentioned previously, arrestors may end up absorbing a lot of energy when limiting surges on a transmission line. If an arrestor dissipates more energy than it is rated as able to handle, it can be damaged or even fail altogether The final factor to consider when selecting an arrestor is the voltage level that the protected component can withstand Clearly, the arrestor should limit the voltage level it allows to an acceptable level for an acceptable period of time [5]. Once these four characteristics of the application of the arrestor are used to pick an arrestor, the protective level of the arrestor must be correlated with the insulation. While the arrestor does limit the voltage across the protected component to a certain discharge voltage, there still must be sufficient insulation to ensure that even at this relatively low voltage level flashover does not occur. The process of matching the characteristics of the insulation equipment with the characteristics of the protective device is referred to as insulation coordination. The discharge voltage of the arrestor depends on the waveshape and rise time of the applied voltage surge. To determine the enough insulation is provided in order to withstand a lightning strike, typically a 10 kA, 8 x 20 µsec discharge is applied experimentally. If the insulation is sufficient, the arrestor will not flash over during the experiment. Other experiments are run to ensure that the insulation is sufficient to handle other situations [1]. The protective margin is a quantity which summarizes the safety of the insulation selection. It is the difference between the insulation withstand voltage and the arrestor ceiling voltage divided by the arrestor ceiling voltage. A safe system will have a relatively large protective margin. A margin of 40% is appropriate unless all uncertainties are evaluated [5]. It is important to appreciate one of the main trade-offs involved in arrestor selection. It is desirable to limit the voltage in question to a relatively low level because this makes the insulation coordination easier. Unfortunately, as the discharge level of the arrestor decreases, the leakage current during steady state operation increases. This leads to larger unwanted power dissipation and higher costs. The trade-off between these costs and the savings resulting from lower insulation coordination costs must be weighed carefully when choosing an arrestor. To communicate the principles of arrestor selection more thoroughly, an arrestor will be selected for a hypothetical application. In this application, surge arrestors are to be applied to the line-to-ground terminals of a circuit breaker with a 38 kV rating and a BIL of 150 kV used on a grounded 34.5 kV system. The highest expected continuous system voltage is 37 kV, but the phase-to-ground voltage can rise to 27.9 kV rms during fault conditions, and faults can persist for up to 20 cycles. The MCOV rating of the chosen arrestor must meet or exceed 37/√3 = 21.4 kV rms, the maximum sustained line-toground voltage level. The TOV rating of the arrestor must be greater than 27.9 kV rms, the expected TOV level of the system [5]. One possible arrestor that would meet these criteria is the Fig. 10. Voltage between the high and neutral lines over time in the PSCAD simulation. 27 NotekV rms polymer how the voltage Electric. A picture of is limited to 12 kV (2.4 time the operating voltage) after just 4 µsec. 10 intermediate arrestor made by General this arrestor is shown in Fig. 18. [4] [5] [6] [7] [8] [9] Fig. 18. GE 27 kV rms polymer intermediate arrestor [8]. This arrestor has a MCOV of 22 kV rms, which exceeds the required value of 21.4 kV rms. This arrestor has is rated for a TOV of 34.5 kV rms, so it fulfills the requirements on TOV (must be greater than 27.9 kV rms). Finally, the insulation level must be considered. In this case the protective margin can be calculated by comparing the 10 kA discharge level of the arrestor (72 kV rms) and the insulation BIL (150 kV rms). This leads to a very safe protective margin of 108%. This surge arrestor would be an excellent choice for this application [8]. V. CONCLUSION While lightning strikes can be incredibly damaging to power systems, that does not mean that they have to be. If the appropriate surge arrestor is selected and combined with the appropriate level of insulation, most lightning strikes will not damage the system. Future work in this field should be focused on finding materials with better nonlinear characteristics and more distinct discharge voltages. This will allow arrestors to protect equipment with less insulation and without allowing too much current leakage during steady state operation. ACKNOWLEDGMENTS M. J. Bloem would like to thank Professor Paulo Ribeiro of Calvin College in Grand Rapids, Michigan for his advice and guidance during the research for and writing of this paper. REFERENCES [1] [2] [3] Spezia, Carl J. and Constantine I. Hatziadoniu. Electrical Power Systems Engineering [E-book]. Ch. 3, “Electrical Transients.” De la Rosa, Fransisco. “Characteristics of Lightning Strokes” in The Electrical Power Engineering Handbook. L. L. Grigsby, Ed. Boca Raton, FL: CRC Press in cooperation with IEEE Press, 2001, pp. 10-2 – 10-7. Chowduri, Pritndra. “Overvoltages Caused by Direct Lightning Strokes” in The Electrical Power Engineering Handbook. L. L. Grigsby, Ed. Boca Raton, FL: CRC Press in cooperation with IEEE Press, 2001, pp. 10-8 – 10-20. Chowduri, Pritndra. “Overvoltages Caused by Indirect Lightning Strokes” in The Electrical Power Engineering Handbook. L. L. Grigsby, Ed. Boca Raton, FL: CRC Press in cooperation with IEEE Press, 2001, pp. 10-21 – 10-35. Lambert, Stephen R. “Insulation Coordination” in The Electrical Power Engineering Handbook. L. L. Grigsby, Ed. Boca Raton, FL: CRC Press in cooperation with IEEE Press, 2001, pp. 10-93 – 10-101. Manitoba HVDC Research Centre, Inc. PSCAD/EMTDC Ver. 3.0.8, 2001. Greenwood, Allan. Electrical Transients in Power Systems, 2nd ed. New York: John Wiley & Sons, Inc., 1991, ch. 16.4 “Surge Suppressors and Lightning Arrestors.” TRANQUELL ® Surge ArrestersProduct Selection & Application Guide. General Electric Capacitor and Power Quality Products, Ft. Edward, NY, 1999. Mathsoft Engineering and Education, Inc. MathCad Ver. 11.0a, 2002. Michael J. Bloem (M ’03) was born in Detroit, MI on February 8, 1982. Michael received his secondary education from the International School Manila and Singapore American School. Michael currently attends Calvin College in Grand Rapids, MI, where he is working towards a BSE with an electrical and computer engineering concentration and a second major in economics. He will receive this degree in May of 2004. He currently works as an Engineering Intern for Smiths Aerospace in Grand Rapids, MI. Prior to working with Smiths, he was an Engineering Intern at Delphi Automotive. During the summer of 2002, he worked as a Research Assistant on a project involving the mathematical modeling of forest growth at Calvin College with Professors Rikki Wagstrom and Randy Van Dragt. He is also currently working on a senior design project involving three-dimensional motion capture using inertial sensors.