i

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 2, MARCWAPRIL 1993

391

Voltage Sags in Industrial Systems
Mark

F. McGranaghan, Member, IEEE, David R.

Mueller, Member, IEEE, and Marek J. Samotyj, Member, IEEE

Abstract-This paper describes the causes of voltage sags in industrial plants, their impacts on equipment operation, and possible solutions. The definition proposed focuses on system faults as the major cause of voltage sags. The sensitivity of different types of industrial equipment, including adjustable speed drive controls, programmable logic controllers, and motor contactors, is analyzed. Available methods of power conditioning for this sensitive equipment are also described.

2

2’
> -1

io
-0

I. INTRODUCTION VOLTAGE SAG is a momentary (i.e., 0.5-30 cycles) decrease in the rms voltage magnitude, usually caused by a remote fault somewhere on the power system (Fig. 1). Voltage sags are the most important power quality problem facing many industrial customers. Equipment used in modem industrial plants (process controllers, programmable logic controllers, adjustable speed drives, robotics) is actually becoming more sensitive to voltage sags as the complexity of the equipment increases. Even relays and contactors in motor starters can be sensitive to voltage sags, resulting in shutdown of a process when they drop out. It is important to understand the difference between an interruption (complete loss of voltage) and a voltage sag. Interruptions occur when a protective device actually interrupts the circuit serving a particular customer. This will normally only occur if there is a fault on that circuit. Voltage sags occur during the period of a fault for faults over a wide part of the power system. Faults on parallel feeder circuits or on the transmission system will cause voltage sags but will not result in actual interruptions. Therefore, voltage sags are much more frequent than interruptions. If equipment is sensitive to these voltage sags, the frequency of problems will be much greater than if the equipment were only sensitive to interruptions. This paper describes the voltage sag characteristics and the sensitivity of equipment. With this information, the range of fault locations on the power system that can cause problems can be estimated (area of vulnerability). Options for improving equipment performance in the presence of voltage sags include power conditioning or equipment design modifications. Both of these options are described. 11. CAUSES OF VOLTAGE SAGS Voltage sags are typically caused by fault conditions. Motor
Paper ICPSD 91-51, approved by the Power Systems Protection Committee of the IEEE Industry Applications Society for presentation at the 1991 Industrial and Commercial Power Systems Department Technical Conference, Memphis, TN, May 6 9 . Manuscript released for publication April 29, 1992. M. McGranaghan and D. Mueller are with Electrotek Concepts, Inc., Knoxville, TN 37932. M. Samotyj is with Electric Power Research Institute, Palo Alto, CA 94303. IEEE Log Number 9207100.

A

Fig. 1. Voltage sag waveform caused by a remote fault condition (six cycles).

Substatlon

Fig. 2. Typical distribution system one line diagram illustrating types of protection devices.

starting can also result in undervoltages, but these are typically longer in duration than 30 cycles and the associated voltage magnitudes are not as low. Motor starting voltage variations are often referred to as “voltage flicker,” especially if the motor starting can occur frequently. Faults resulting in voltage sags can occur within the plant or on the utility system. The voltage sag condition lasts until the fault is cleared by a protective device. In the plant, this will typically be a fuse or a plant feeder breaker. On the utility system, the fault could be cleared by a branch fuse or a substation breaker. If reclosing is used by the utility, the voltage sag condition can occur multiple times. Utility system faults can occur on the distribution system or on the transmission system. Fig. 2 illustrates a typical distribution system configuration with a number of feeders supplied from a common bus. A fault on Feeder F1 will cause an interruption which will affect the customers on that feeder. However, all of the customers on the three parallel feeders will experience a voltage sag while the fault is actually on the system. With the reclosing breakers at the substation, the customers on parallel feeders can experience as many as four

0093-9994/93$03.00 0 1993 IEEE

I
398
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 2, MARCWAPRIL 1993

\Recloser Operalinq Sequenceb

I

Conlocls Closed

\

I
Reclosing Intervals (Contacts Open)

I

Fig. 3. Typical recloser operating sequence.

2oov

,

7

ov

20 0

V/div

verticol

2

sec/div

horiz

Fig. 4. Voltage sag at customer location with one recloser operation.

voltage sags in succession, lasting for durations ranging from a couple of cycles to more than ten cycles (see typical reclosing sequence in Fig. 3). The voltage characteristic at the customer location on a parallel feeder will look something like the plot in Fig. 4 (one reclosing operation). Faults on the transmission system can affect even more customers. Customers hundreds of miles from the fault location can still experience a voltage sag resulting in equipment misoperation when the fault is on the transmission system. The large majority of faults on a utility system are single line-to-ground faults (SLGF). Three phase faults are more severe, but much less common. SLGF’s often result from weather conditions such as lightning, wind, and ice. Contamination of insulators, animal contact, and accidents involving construction or transportation activities also cause faults. Although utilities go to great lengths to prevent faults on the system, they cannot be eliminated completely. Lightning is the most common cause of faults on overhead transmission and distribution lines. Lightning can cause a fault by directly striking a phase conductor (direct strike) or by striking a grounded object, such as a shield wire or tower (backflash). In either case, the voltage developed across the phase conductor insulators close to the stroke location can cause a flashover which then results in the flow of fault current. Usually, these faults are temporary, which means that they

will not reinitiate after they have been cleared and the line is reclosed. The probability of a flashover during a lightning stroke can be reduced by applying surge arresters to divert the lightning current to ground. The probability of backflashes can be reduced by minimizing footing resistances. Some utilities are considering the application of arresters along transmission lines to reduce the incidence of lightning-induced faults on transmission systems. Regardless of the measures taken, lightning-induced faults cannot be eliminated completely. There will still be lightning strokes with high current magnitudes and rates-of-rise which will cause flashovers. Therefore, it is important for customers to make sure that critical equipment sensitive to voltage sags is adequately protected.
1 1 RANT 1. VOLTAGE DURING SINGLE ON LINE-TO-GROUND FAULTS THE UTILITY SYSTEM

Single line-to-ground faults (SLGF’s) on the utility system are the most common cause of voltage sags in an industrial plant. The voltage on the faulted phase goes to zero at the fault location. The voltage at the substation and on parallel feeders will depend on the distance of the fault from the substation. On transmission systems, the faulted phase voltage at a remote location depends on the overall network impedances. The important quantities for equipment sensitivity are the voltages at the customer bus. These voltages will depend on the transformer connections between the faulted system and the customer bus. For a distribution system fault, the worst case occurs when the fault is close to the substation bus. Effectively, this is the same as a fault near the customer transformer primary (Fig. 5). The voltages on the customer bus will then be a function of the customer transformer connections, as indicated in Table I. The relationships in Table I are very important. One might think that an SLGF on the primary of a wye groundeddelta transformer could result in zero voltage across one of the secondary windings. Instead, circulating fault current in the

I
MCGRANAGHAN: VOLTAGE SAGS IN INDUSTRIAL SYSTEMS 399

C

480 Vok

U t i l i t y Co. Source

Transformer
Connection
J
Motor-Acmss Ute unc Conbol

lighls. solenoids

Fig. 5. Single line-to-ground fault near substation.

Adjuotablc Spccd Drlvc

TABLE I TRANSFORMERSECONDARY VOLTAGES AN SLGF ON THE PRIMARY WITH
Phase t Phase o
Vab

Fig. 6 . Typical single-phase and three-phase loads.

Vbc

Vca

Jab

vbc

Vca

I v . SENSITIVITY OF EQUIPMENT VOLTAGE SAGS TO Industrial plant power is often distributed by three-phase 480 V feeders. The loads can be categorized by type and connection to the power system (Fig. 6): Motors, heating elements, and other three-phase loads can 0.58 1.00 0.58 be connected directly to the 480-V feeders. Adjustable-speed drive and other power electronic devices that use three-phase power will be connected directly to the 480-V feeders or through an isolation transformer. 0.33 0.88 0.88 Lighting often utilizes single-phase 277-V connections from phase-to-neutral or may use 480- or 120-V singlephase connections. I Control devices such as computers, contactors, and programmable logic controllers utilize 4801120-V single0.88 0.88 0.33 phase transformers for 120-V control. The voltages experienced during a voltage sag condition will depend on the equipment connection. Table I showed that the individual phase voltages and phase-to-phase voltages are delta secondary windings results in a voltage on each winding. quite different during an SLGF condition on the transformer The magnitude of the lowest secondary voltage depends upon primary. Some single phase loads will be unaffected and other single phase loads may drop out, even though their sensitivities the relationship to voltage sags may be identical. Voltage unbalance is also a concern for motor heating. However, the durations of the unbalanced voltages during fault conditions are so short that motor heating is not a significant concern. Different categories of equipment and even different where brands of equipment within a category (e.g., two different models of adjustable speed drives) have significantly different X , transformer short circuit reactance sensitivities to voltage sags. This makes it difficult to develop a single standard that defines the sensitivity of industrial process equipment. XS source equivalent reactance. The closest document to a standard is the CBEMA curve given in Fig. 7, which was developed by the Computer For industrial power distribution, the ratio a will usually be Business Manufacturers Association [3]. This applies primarily very close to unity and the relationships in Table I are for to data processing equipment. The curve shows that the load this case. sensitivity is very dependent on the duration of the sag. Even with an SLGF on the primary of the transformer, the Allowable sags range from 0% voltage for 1/2 cycle to only voltage sag at the customer bus will be no lower than 33% 87% voltage for 30 cycles. normul value. These faults account for the greater majority of While the CBEMA limits suggest a "standard" sensitivity faults on the power system. to voltage sags, actual plant equipment has a variety of op0.58
1.00 0.58

I

I
400
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 2, MARCWAPRIL 1993

CBEMA
400

300

voltage in 200
Percent

100

0 0.001

0.01

0.1

0.5 1.0

6 10

30
2s

1000

Time in Cycles (60 Hz)

ANSIIIEEE - Std 446-1987 Typical Design Goals of Power-Conscious Computer Manufacturers
Fig. 7. CBEMA operating voltage envelope.

erational characteristics during voltage sags. A few examples are listed here. I) Motor Contactors and Electromechanical Relays: One manufacturer has provided data that indicates their line of motor contactors will drop out at 50% voltage if the condition lasts for longer than one cycle. This data should be expected to vary among manufacturers, and some contactors can drop out at 70% normal voltage or even higher [4]. 2) High-Intensity Discharge (HID) Lamps: Mercury lamps are extinguished at around 80% normal voltage and require time to restrike [ 5 ] . A voltage sag that extinguishes HID lighting is often mistaken as a longer outage by plant personnel. 3) Adjustable Speed Motor Drives (ASD’s): Some drives are designed to ride through voltage sags. The ride through time can be anywhere from 0.05 to 0.5 s, obviously depending on the manufacturer and model. Some models of one manufacturer monitor the ac line and trip after a voltage sag to 90% normal is detected for 50 ms. 4 ) Programmable Logic Controllers (PLC’s): This is an important category of equipment for industrial processes because the entire process is often under the control of these devices. The sensitivity to voltage sags varies greatly but portions of an overall PLC system have been found to be very sensitive. The remote VO units, for instance, have been found to trip for voltages as high as 90% for a few cycles [6]. The sensitivity range for these types of equipment is shown in Fig. 8 with the durations of fault induced voltage sags also indicated. The wide range of sensitivities underlines the importance of working with the manufacturer to make sure the equipment can work in the environment where it will be used and to develop specifications based on realistic power system conditions. It is important to recognize that the entire process in an industrial plant can depend on the sensitivity of a single piece of equipment. The overall process involves controls, drives,

100 90 VOLTAGE 80

(PERCENT 70

NORMAL)

: :
40 30 20 10
0

0

10 TIME (CYCLES)

20

30

Fig. 8. Range of equipment sensitivity to voltage sags.

motor contactors, robotics, etc., that are all integral to the plant operation. This can also make it difficult to identify the sensitive piece of equipment after the entire process shuts down.

v.

ESTIMATING PROBABILITY THE

OF A VOLTAGE SAG PROBLEM

The most frequent cause of voltage sags at a large industrial plant is lightning. Lightning is weather related, and the weather can be extremely variable from one season to another, or one year to another. But over longer periods of time, weather will more closely follow certain pattems. Activities such as those by the National Lightning Detection Network are establishing the amount of lightning strokes a given area will receive over longer periods of time. The results of this paper report on ground flash density (Ng) for all areas of the country. The ground flash density is a measure of lightning strokes to ground per square km per year. It is more accurate than the previously used isokeraunic level in determining the expected lightning performance of transmission lines. Isokeraunic level is the number of days

I
MCGRANAGHAN: VOLTAGE SAGS IN INDUSTRIAL SYSTEMS
40 1

Plant Voltage vs. Fault Location
100
90 Normal Volraoa

1W1 1%

Area of Vulnerabilib for Senoithn Loads Area of Vulnerabilily for Motor Contactors

T

p - L - t -

0

2600

6000

76W

loo00

12600

16ooO

Distance From Substationto Faun (Feet)
I

Fig. 9. Voltage sag versus fault location.

I

0 1 2 3 4 5

m

Miles

per year lightning is heard, and must be multiplied by a proportionality factor to convert it to ground flash density. Utilizing geometry of the transmission lines, BIL levels of the insulators, and ground flash density, the expected number of faults per km of line per year can be calculated [ 2 ] . Fortunately not every fault on the utilities’ transmission grid will disrupt production at the industrial plant. The effect of a fault on plant equipment depends on the fault clearing time and the location of the fault. Computer calculations can determine the plant voltage versus fault location on a distribution system and curves similar to those in Fig. 9 can be used to illustrate the range of fault locations that can cause problems. Factors governing the magnitude and duration of voltage sags include the fault impedance and location, the configuration of the power network, and the system protective relay design. Fig. 2 illustrates a relationship between fault location and voltage sag magnitude, for an example distribution system (through a wye-delta transformer connection). Fault clearing time is dependent on the utility company’s system protection practices. If instantaneous fault clearing is used, 4-6 cycles is a likely duration of the voltage sag. If an intentional time delay is used, the duration might be as long as 15 cycles. On transmission systems, it is more difficult to determine the range of fault locations that can result in unacceptable voltage sags. Computer simulations can be used to determine voltages around the system for any fault location. These calculations can be used to define an “area of vulnerability” for a particular customer (Fig. lo). The likelihood of a fault within this area can then be calculated. SOLUTIONS TO VOLTAGE SAG PROBLEMS VI. EVALUATING The interruption of an industrial process due to a voltage sag can result in very substantial costs to the operation. These costs include lost productivity, labor costs for clean-up and restart, damaged product, reduced product quality, delays in delivery, and reduced customer satisfaction. Proper evaluation of alternatives to improve plant equipment and the power distribution network requires a cost versus benefit comparison. For example, once the costs of retrofitting sensitive process equipment with some method of improving voltage sag ride through are determined, the benefits

Fig. 10. Area of vulnerability to transmission system faults for a particular customer location.

of recovering lost production, material, product quality, and customer responsiveness must be determined. Experience by the industrial plant will provide data on production losses for a given occurrence following a voltage sag. There may even be a record the number of disruptions due to voltage sags in the past calendar months or years. If the necessary data exists, the cost of implementing a solution can be evaluated against the expected cash flow of recovered production losses. Solutions to the voltage sag problem must almost always be implemented in the customer facility. As mentioned previously, it is possible for the utility to reduce the number of faults on the system through design practices and additional equipment, but it is never possible to eliminate faults on the system. The plant equipment must be designed to handle the most common voltage sag conditions or be retrofitted with appropriate power conditioning. In the long run, the best solution to voltage sag problems will be to purchase equipment that has the necessary ride through capability. As manufacturers become increasingly aware of the need for this capability, it will become more and more standard in industrial process equipment. Even now, manufacturers offer new models or simple modifications that permit extended ride through capability. Until equipment can handle voltage sags directly, it will often be necessary to apply power conditioning equipment for particular sensitive loads. Most voltage sag conditions can be handled by ferroresonant, or constant voltage, transformers (CVT’s). CVT’s are especially attractive for loads with relatively low power requirements and loads which are constant. Variable loads are more of a problem for CVT’s because of the tuned circuit on the output. These power conditioners work similar to a transformer being excited high on its saturation curve, so that the output voltage is not significantly affected by input voltage variations. The actual design and construction is more complicated. A typical ferroresonant circuit is shown in Fig. 11. Ferroresonant transformers output over 90% normal voltage as long as the input voltage is above a minimum value, at which the output collapses to zero voltage. Table I1 shows

402

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 2, MARCWAPRIL 1993

Windinn

1
I

Ncutrrlizing Winding

Fig. 1 1. Typical circuit for a ferroresonant transformer.

performance of transmission and distribution overhead lines can predict the frequency of lightning induced faults, or be compared to existing performance. 3) A single line-to-ground fault on the primary side of a distribution transformer will result in a voltage sag to no lower than 33% of normal voltage on any phase-to-phase connection. 4) The sensitivity of industrial equipment to voltage sags varies greatly. The more sensitive equipment widens a plant’s area of vulnerability to disruptive voltage sags. 5 ) Constant Voltage Transformers can be applied economically at constant loads to handle the great majority of voltage sag conditions. If needed, increased protection for voltage sags or actual interruptions can be provided in the form of UPS systems. REFERENCES
[ I ] L. Tang and M. McGranaghan, “Power quality definitions and standard terms development,” EPRI, Res. Project RP3098- 1, June 1990. [2] J. G. Anderson, “Lightning performance in transmission lines,” in EPRI Transmission Line Reference Book, pp. 545-597. 131 Recommended Practice for Emergency and Standby Power for Industrial and Commercial Applications, IEEE Standard 446-1987. [4] D. M. Sauter, “Voltage fluctuations on power systems,” in Westinghouse Electric Utili@ Engineering Reference Book, Distribution Systems, 1965, p. 362. [ 5 ] Recommended Practice for Electric Power Distribution for Industrial Plants, ANSI-IEEE Standard 141-1986. [6] V. E. Wagner, A. A. Andreshak, and J. P. Staniak, “Power quality and factory automation,” IAS Annu. Meeting, vol. 35, no. 6, pp. 1391-1396.

TABLE I1 ALLOWABLE VOLTAGE SAG VERSUS TRANSFORMER LOADING FERRORESONANT FOR TRANSFORMERS
~~ ~

Amount of Load 1/4 of rating 1/2 of rating Full loaded

Minimum Input Voltage” 36 V (30%’of 120 V) 55 V (46% of 120 V)

85 V (71% of 120 V) Voltage collapses, even at full 150% of rating (overloaded)* input aIf the input voltage is above the minimum input voltage, the output voltage will remain in the range of +3 to -6% of normal value. These transformers are also available with 480-V input. bInrush current greater than 150% rating will cause the voltage to collapse. Inductive loads such as contactors require special consideration to ensure that the transformer has adequate capacity to handle current inrush rwuirements.

that the minimum input voltage is a function of the load. It is noteworthy that at 1/4 load, the ferroresonant transformer will output the necessary voltage even during voltage sags to as low as 30%. This is important since it is virtually impossible for an SLGF condition on the utility system to cause a voltage sag below 30% at the customer bus when the customer is supplied through a delta-wye or wye-delta transformer. CVT’s will handle the majority of voltage sag conditions. Voltage sags which are too severe for CVT’s, such as due to a three phase fault close to the customer location, or actual interruptions, could still cause process disruptions. Protection for extremely critical loads, such as life safety systems and critical data processing equipment, should include UPS systems or the equivalent for complete backup capability. VII. CONCLUSIONS

1) Voltage sags are becoming an increasing concern of industrial plants due to increasing automation. Automated facilities are more difficult to restart, and the electronic controllers used are sometimes more sensitive to voltage sags than other loads. 2) Single line-to-ground faults on the utility distribution or transmission system are often the cause of voltage sags. Lighming is a frequent cause. Evaluation of the lightning

Mark F. McGranaghan (M’78) received the B.S.E.E. and M.S.E.E. degrees from the University of Toledo in 1977 and 1978, respectively, and the M.B.A. degree from the University of Pittsburgh in 1985. He is responsible for development and marketing of power systems products at Electrotek Concepts Inc., Knoxville, TN. These products include hardware and software for power quality analysis, harmonic analysis, and transient analysis on power systems. In addition, he manages and performs studies in the areas of market assessment and power quality analysis. He has been involved in a wide range of activities dealing with concems for power quality and the integration of power electronic equipment on the power system. He has developed and taught harmonics, transients, and power quality seminars during the past 10 years, and has been actively involved in rewriting the IEEE Harmonics Standard (5 19.1981). Before joining Electrotek, he worked in the Systems Engineering Group at McGraw-Edison. There, he directed work on engineering studies employing both the Transient Network Analyzer (TNA) and digital computer programs. He has directed studies for utilities and consultants throughout the United States and in many foreign countries. His primary study areas were transient analysis, insulation coordination, harmonic analysis, series capacitor protection, shunt capacitor switching, SVS applications, flicker analysis, and equipment failure studies. Mr. McGranaghan is a member of the IEEE Power Engineering Society and is actively involved in a number of IEEE and CIGRE committees. He has been chairman of the Switching Surge Working Group and is Chairman of ANSI-C92 on Insulation Coordination. Currently he is working on revising IEEE 519-1981 (Harmonic Standard) and is a liaison to the EEI Working Group on Power Quality.

I
MCGRANAGHAN: VOLTAGE SAGS IN INDUSTRIAL SYSTEMS
403

David R. Mueller (M’91) received the B.S.E.E. degree from the University of Cincinnati in 1982 and the M.Eng. degree in electric power engineering from Rensselaer Polytechnic Institute in 1990. He is employed with Electrotek Concepts, Knoxville, TN, as a Senior Power Systems Engineer. He has primarily been responsible for power quality investigations at industrial plants. He has been responsible for studies to identify the causes of power system equipment failure, or the misoperation of process controls due to power disturbances. He has also conducted several studies involving power factor correction and harmonic filtering. He has taught several power quality seminars. Prior to joining Electrotek, he was employed for eight years with Delco Products Division of General Motors. During his employment with GM, his assignments included the design and troubleshooting of machine controls, maintenance, and upgrade of the plant power distribution system, and energy conservation activities. Mr. Mueller is a member of the IEEE Industry Applications Society and is a Registered Professional Engineer.

Marek J. Samotyj (M’88) received the B.S. and M.S. degrees in electrical engineering from Silesian Polytechnical University, Gliwice, Poland, in 1969 and 1971, respectively, and the M.S. degree in engineering-economic systems from Stanford University in 1985. He is Manager of Power Electronics End-Use Systems in the Power Electronics and Controls Program, of the Customer Systems Division at the Electric Power Research Institute (EPRI) in Palo Alto. CA. During 1984. he was a Consultant for Power Electronic Systems at EPRI; he joined the Institute in 1985. He is responsible for the applications and field testing of adjustable speed drives, and conducts research in the areas of power quality and end-use magnetic fields management. Before coming to EPRI, he was a Research Assistant for the Energy Modeling Forum at Stanford University (1982-1985). From 1980 to 1981, he was a Consulting Member of the Scientific Board of the Future Research Center at Technical University in Wroclaw, Poland, and a Consultant for R&D planning strategy with the Commission on the National Economic Reform in Warsaw, Poland. From 1971 to 1975, he was a Consulting Staff Engineer and Project Manager for the Polish Ministry of Mining and Energy. Mr. Samotyj is a member of CIGRE, which is the International Conference on Large High-Voltage Electric Systems. Currently he is a secretary of the IEEE Power Quality Standards Coordinating Committee 22. He also actively participates in C E R E Working Groups on adjustable-speed drives and on power quality. From 1981 to 1982, he was a Fulbright Senior Scholar and the Fellow of the Professional Journalism Program at Stanford University.