CHAPTER

22

LAMP FLICKER ON POWER SYSTEMS

Original Author:
S. B. Griscom OLTAGE regulation has been one of the most important problems of the electric industry since its inception. The sizes of many parts of a power system are determined largely by this one consideration alone. A large proportion of the selling price of electrical power is the interest and other fixed charges on production and distribution facilities, so that any improvement in regulation is ultimately reflected in higher rates. Similarly, types of load imposing exceptionally severe regulation requirements will also increase the cost of supplying energy. In the early days of the industry, a relatively wide range of voltage variation was permissible, because the public was at that time unaccustomed to uniform lighting intensity. Today, there is a greater consciousness as to whether the voltage level is about right, as indicated by the whiteness of the light and by lamp life. While, however, a narrower voltage band is required than formerly, this is not always the limiting factor in voltage regulation. Numerous new devices have been added to power lines in the last few years, which impose rapid and frequent changes of load, with correspondingly rapid voltage changes. Repeated observations have shown that rapid changes of voltage are much more annoying than slow ones, so that flicker effects may limit the useful loadcarrying ability of individual circuits long before maximum steady-state regulation or heating is reached. Consequently, the voltage regulation problem must now be considered from two angles: the normal drop in voltage from light load to full load, and the superimposed flickers due to motor-starting and to various pulsating and irregular loads. The differences in voltage between light and full load affect the performance, efficiency, and life of electrical equipment, and are treated in Chap. 10. The present chapter considers only the flicker component of voltage regulation, and deals primarily with the reaction of the human eye to variations in electric light intensity. I. PERMISSIBLE FLICKER

Revised by:
S. B. Griscom change, duration of change, and frequencyof occurrence of the flicker. These and other factors greatly complicate the problem of assigning limits to permissible flicker voltages. Numerous investigators have studied the flicker problem. The most complete analysis is found in the report The Visual Perception and Tolerance of Flicker, prepared by Utilities Coordinated Research, Inc. and printed in 1937, from which Figs. 1 to 4 of this chapter are reproduced. Figure 1 shows the cyclic pulsation of voltage at which flicker of 115 volt tungsten-filament lamp is just percep-

Fig. lCyclic pulsation of voltage at which flicker of 115-volt tungsten filament lamp is just perceptible-derived from 1104 observations by 95 persons in field tests of 25-watt, 40-watt, and 60-watt lamps conducted by Commonwealth Edison Company. Figures on curves denote percentages of observers expected to perceive flicker when cyclic voltage pulsations of indicated values and frequencies are impressed on lighting circuits. Plotted points denote medians of observationat various frequencies, number of observations in each case being indicated by adjacent figures.

The permissible amount of flicker voltage cannot be stated concisely for several reasons. There is first the human element; one individual may think objectionable a flicker not perceptible to another. The lighting fixture used is of considerable importance. Smaller wattage incandescent lamps change illumination more quickly upon a change of voltage than lamps with heavier filaments. The character of the voltage change is also important. Cyclic or rapidly recurring voltage changes are generally more objectionable than non-cyclic. On non-cyclic changes the annoyance due to the flicker is affected by the rate of

tible. Flickers as low as g volt were perceptible in 10 percent of the observations, when the rate of variation was 8 cycles per second. In order for the variations to be perceptible in 90 percent of the observations, however, the voltage change had to be over one volt at the same frequency. The range between 6 and 12 cycles per second was the most critical. Figure 2 shows the minimum abrupt voltage dip to cause perceptible flicker in a 60-watt, 120-volt tungsten-filament lamp, as a function of intensity of illumination. Curves are shown for 5 and 15 cycles (60 cycles per second basis) durations of voltage dip. It should be noted that abrupt voltage dips of 1.5 to 2.0 volts were perceptible.

719

Lamp

Flicker

on Power

Systems

Chapter 22

Fig. 2Minimum abrupt voltage drop for perception of flicker of 60-watt, 120-volt coiled-coil tungsten-filament lamps operated on 60-cycle alternating current. Each point represents the means of the observation of 44 persons.

Fig. 4Recommended

maximum allowable voltage.

cyclic variation of

Figure 3 shows the effect of duration of transition of voltage on the average threshold of perceptibility of flicker for tungsten-filament lamps. This curve shows quite clearly that whereas an abrupt change of about 1% volts

Fig. 3Effect of duration of transition of voltage on average threshold of perceptibility of flicker of tungsten-filament lamps.

is perceptible, a change of 5 volts or more is necessary before voltage variations requiring several seconds for completion can be perceived. Figures 1 to 3 are of interest in showing the perceptibilities for various classes of flicker voltages. These are not working limits, because a perceptible flicker is not necessarily an objectionable one. Fig. 4 shows the recommended maximum allowable cyclic variation of voltages as set up by various authorities for their own use. The variations in these recommendations is an indication of the extent to which individual judgment enters the problem. The curves are nevertheless an exceedingly valuable guide. Cyclic flicker, when perceptible, is likely to be objectionable, at least to some individuals. Isolated voltage dips, however, even if plainly perceptible, are not objectionable to the majority of individuals unless rather frequent. It can, therefore, be expected that larger vari-

ations are permissible for non-cyclic than for cyclic variations, but that the amount of tolerable dip depends upon

Chapter

22

Lamp

Flicker

on Power

Systems

721

the frequency of occurrence and the class of service. Here again, judgment is an important factor as well as technical facts. The maximum allowable fluctuations practiced by one operating company are shown in Table I. This is a very comprehensive set of standards and has proved satisfactory in practice.

II. ORIGIN OF FLICKER VOLTAGES

Flicker voltages may originate in the power system, but most frequently in the equipment connected to it.

1. Generating Equipment Prime Mowers-Engine driven generators are probably responsible for most of the rare cases of flicker originating due to the power system itself. Curve (a) of Fig. 5 shows the variation in tangential force of a four cylinder

Fig. 5Curves from a four-cylinder 300 rpm Diesel engine at full load driving a generator. The variation in velocity caused a corresponding variation in the generated voltage.

300 rpm Diesel engine at full load, and Curve (b) shows the corresponding percent change in angular velocity of the rotating parts. With all other factors constant, this non-uniform rate of rotation produces a fluctuation in amplitude of the generator voltage The total variation in voltage is the same as the total variation in speed; in this example 0.7 percent. The frequency of the variation is equal to the rpm times the number of power strokes per revolution; in this case 300x2 = 600 per minute or 10 per second. Referring to Fig. 1, it is seen that 0.7 percent change in voltage is readily perceived by most individuals. Fig. 4 indicates that most operators regard this as too much flicker to be tolerable. About the only practicable remedies are increasing the flywheel effect, or changing the speed to get within a less objectionable frequency range. In this actual case, the flicker of the original installation caused many complaints and it was satisfactorily corrected by increasing the flywheel effect. When two or more engine-driven generators are in continuous operation at the same station, the amplitude of the fluctuation can frequently be lowered, and the fre-

quency doubled to get it out of the objectionable range, by synchronizing the generators so that the power strokes of the two engines alternate rather than occur simultaneously. This can be done because there are usually more poles on the generators than cylinders on the engine, particularly in those engines where the flicker is in the objectionable range. A stroboscope or similar device used with the regular synchroscope permits such synchronizing. It has sometimes been thought that it should be possible to correct flicker of this type by the use of special voltage regulators of unusually fast response. In practically every case this is completely out of the question because the frequency of the flicker is too high for the time constant of the generator field. For example, the field time constant of a typical moderate-sized engine-type generator is between 0.5 and 1.5 seconds, whereas the range of most objectionable flicker is between 1/4and $ second per cycle. Even electronic excitation systems are unable to regulate voltage at such a high rate. Generators-A symmetrical generator with constant load, excitation and angular velocity produces a constant terminal voltage. If any of these quantities varies, however, the terminal voltage also varies. It is possible to have a sufficient degree of non-uniformity in the generator air gap to cause pulsating terminal voltage. However, the commercial manufacturing tolerances are sufficiently close that no case of flicker due to this cause is known to have occurred. To produce flicker in this manner, both the rotor and stator must be eccentric. Stator bores of all but the smallest size machines will inherently have a certain degree of eccentricity, because they must be built up with segmental laminations. In spite of this built-up construction, quite close tolerances are held by the use of accurate dies and assembly keys and dowels. Further attempts at improvement would be very difficult as it would require boring or grinding the inner bore of the stator punchings. This is quite undesirable from the standpoint of accumulation of iron chippings and filings between laminations and into the slots, which might result in a condition of insulation breakdown and localized heating of the stator The rotor eccentricity is, because of the necessity of dynamic balancing, held normally to quite close tolerances. Since no voltage fluctuations can be produced if the rotor is concentric with the shaft, no modification of standard manufacturing procedures has ever been necessary from the standpoint of flicker voltages. Abrupt changes of load on generators produce corresponding changes in the terminal voltages. This voltage fluctuation is the result of two factors: the change in speed, and the regulation of the machine. In central station practice it is very unusual for change in speed to be a significant factor. Sudden load increments are usually too small as compared with the total generating capacity to change the speed materially. Even if the speed changes, however, the rate at which the voltage drops is ordinarily so slow, that the effect is imperceptible to the eye (see Fig. 3). A typical voltage-time regulation curve of a large turbine generator, following sudden application of load is shown in Fig. 6. Speed and excitation voltage are assumed constant. Three points on this curve are of especial in-

722

Lamp

Flicker

on Power

Systems

Chapter from synchronous

22 re-

Similarly, point (b) is calculated actance using the relation:

O-b= E, - Ixd
The transition from (a) to (x) and from Cx) to (b) may be calculated by using the appropriate machine time constants. This procedure is more fully described in Chap. 6. From the standpoint of flicker voltage, the following points are of interest. For single load applications more than 10 cycles in duration (on a 60-cycle system), the voltage regulation point (c) of Fig. 6, calculated from the transient reactance, is the determining quantity. Fig. 2 shows that there is little difference in perception lasting from 5 to 15 cycles of voltage drop. In average machines, the subtransient drop is usually about two-thirds of the transient drop. However, after about the first 5 cycles, the voltage drops to the value determined by transient reactance. A further drop in voltage takes place due to the decrement of the field, reaching point (b) on Fig. 6. Usually, this synchronous reactance drop is not more than two or three times the transient reactance drop. Automatic voltage regulators may limit the drop to less than 1% times the transient drop. Reference to Fig. 3 shows that for a transition time of the order required (3 to 10 seconds), the additional voltage drop due to field decrement is not perceptible For load durations less than 5 cycles, it is likely that the regulation as calculated from the subtransient reactance determines the permissible flicker. While the voltage drop at the end of 5 cycles is greater than initially, the transition is gradual and it is doubtful if the eye can discern so small a difference. For load durations between 5 and 10 cycles, it is probable that an average between subtransient and transient reactances should be used to calculate flicker voltages for comparison with perception data similar to those given in Figs. 1 to 3. The proper reactance to be used to calculate the effect of cyclic variations depends upon the frequency of their occurrence. The following range is suggested for generators 5000 kva and above.

Fig. 6Voltage-time following

regulation of a large turbo-generator sudden application of load.

terest. Point (a) is the voltage immediately following the application of load; point (b) is the voltage after the voltage has settled; point (c) is an extrapolation of the curve from (b) back to zero time. Each of these points may be determined closely by the use of the appropriate generator reactance. In fact, the standard definition of the various reactances has been made for this particular use. For a fuller discussion of machine characteristics see Chap. 6. Point (a) is determined by the use of the machine subtransient reactance xd. In the case of an initially unloaded machine, the voltage (O-a) is the vector difference between the no-load voltage and the product of the load current times the subtransient reactance. That is,

O-a = E, - Ixd
The voltage rapidly falls further to a point (2) and at a much lower rate to point (b). The reason may be described approximately as follows. At the instant of load application, the magnetic flux in the air gap remains substantially constant, and the initial drop in voltage is principally that due to reactance of the armature winding. However, the armature currents set up a demagnetizing effect to buck the field flux. The decreasing field flux generates voltages and currents in the field structure, which resist or delay the ultimate change. The induced currents in some parts of the field structure, such as the eddy currents in the pole face, damper windings, or rivets, subside rapidly because of the high resistance of the path, and allow part of the flux to change quickly. In the average machine, about 0.1 second is required for this change. Most of the change of voltage between points (a) and (x) is due to this cause. The majority of the field flux is encircled by the field winding which is of very low resistance, and, therefore, constitutes an effective damper to rapid changes of voltage. The change in voltage from (x) to (b), therefore, constitutes an effective damper to rapid changes of voltage. The change in voltage from (x) to (b) is, therefore, comparatively slow, from 3 to 10 seconds being required for 90 percent of the change to take place in large machines. Point (x) is not directly calculable by using standard machine reactances alone. Point (c), however, can be calculated in the same manner as point (a), except that transient reactance is used. That is

O-c=E,-Ixd

In smaller machines the field time constant may be so short that pulsation frequencies below 2 cycles per second may require the use of synchronous reactance. Excitation Systems-Excitation systems are rarely the cause of flicker voltages in central station practice. In larger generators, field time constants above 3 seconds cause variation in armature voltage to be very gradual no matter how fast the excitation may change. Occasionally, hunting of generator voltage regulators causes wide voltage fluctuations, but this is not a true flicker. On small generators, continuously vibrating regulators occasionally cause a small pulsation of the armature voltage.

Chapter 22

Lamp

Flicker

on Power

Systems

723

Since the alternator field constant is usually too high to permit exciter fluctuations to show up in the alternator terminal voltage, correction of flicker by means of excitation control is not practical. In other words, the amount of generator flicker depends upon its inherent reactance characteristics and cannot be substantially improved by excitation control. Short Circuits and Switching Surges-Short circuit currents, because of their magnitude, produce large voltage drops and attendant flicker. Reduction in the amount of voltage drop is not feasible without major changes in system layout and large expenditures. The duration of the voltage drop can, however, be markedly reduced in a number of cases by the use of high-speed relays and breakers. Flicker due to short circuits occurs so seldom that no special consideration for this purpose alone is necessary. The tendency is toward a gradual reduction in flicker, as system improvements are made for other purposes such as protection of lines against lightning, installation of high speed relays and breakers, etc. These comments apply to networked systems; in radial lines, short circuits produce outages, a distinctly different problem. Line switching rarely produces flicker unless load is picked up or dropped, or lines with large charging currents are switched. Here again, special provisions to reduce flicker are rarely necessary.

2. Utilization

Equipment

Most of the flicker on central station systems is due to the customers utilization equipment. The following are some of the more common types of equipment known to cause flicker. Motor Starting-Probably most of the flicker problems are caused by the starting of motors. For reasons of cost, efficiency, and reliability, commercial general purpose motors require a momentary starting current several times their full load running current, in order to produce sufficient starting torque. Three general classes of motor installations are of importance in the flicker problem. (1) Single phase fractional horsepower motors commonly used in homes and small stores. (2) Integral-horsepower polyphase motors operated from secondary distribution circuits, such as in small shops, large stores and buildings, and recently in a small number of homes for air conditioning. (3) Large integral-horsepower three-phase motors operated from primary lines, mostly by industrial concerns (1) Single phase fractional horsepower motors are manufactured in large quantities, and to maintain this extent of usage, they must continue to be low in cost, economical, rugged and reliable. These requirements have led to several classes of motors depending upon the service, with one class designed specifically for frequent starting with low starting current. This motor is used in great quantities in domestic refrigerators and oil burners, and the g horsepower 110-volt class usually has a locked-rotor starting current of 20 amperes or less. It is not, unduly expensive

to design a distribution system to supply 20 amperes at 110 volts without objectionable lamp flicker. Where single phase 110/220 systems are used, 40-ampere starting currents are permissible on the 220-volt connection, allowing larger motors to be used. (2) Integral-horsepower motors on secondary circuits are potential sources of flicker. In most cases, such motors are used in areas of high load concentration and the power circuits are correspondingly large. This usually permits ordinary 3-phase squirrel cage motors to be started directly across the lines. In some cases, however, the size of a motor is out of proportion with its supply line. The practical solution is to use a starter that limits the initial inrush of current and thereafter changes the current in increments sufficiently small to prevent objectionable lamp flicker. (3) Supplying large motors from primary power lines is usually not troublesome because such motors are usually located in an industrial district where power supply lines are inherently heavy and where wider limits of voltage drop are permissible (See Table 1). There are nevertheless a number of cases particularly in rural communities, where motor ratings are too high for the power facilities. A suitable motor starter may correct such cases, although in some installations other measures may be required. Starting currents for both induction and synchronous motors at full voltage vary from 5 to 10 times full load, depending upon the size, number of poles, and other application requirements, such as required starting, pull-in, and pull-out torques. The power factor under lockedrotor conditions varies between 25 and 50 percent. For approximate calculations, a starting current of 6 times normal at 35 percent power factor may be used. Wide variations from this should be expected, and specific data should always be used when obtainable. Motor-Driven Reciprocating Loads-This type of load usually consists of air compressors, pumps and refrigerators. The motor load varies cyclically with each power stroke and produces a corresponding variation in the line current. Thus, comparatively small variations of voltage may be objectionable if the pulsation occurs 6 to 12 times per second. (See Fig. 1.) Difficulty from this source has been claused in the past by domestic refrigerators, but in modern designs both the frequency of pulsation and the amount of fluctuation have been improved, so that complaints from this cause are now rare. Figure 7 is an oscillogram showing the armature voltage, current and three-phase power of an air compressor driven by a 100-horsepower wound rotor induction motor. There are several points of interest on this oscillogram. First, although the voltage variation can scarcely be detected on the oscillogram, it actually was very objectionable to lighting customers. This shows that oscillographs used in the conventional manner may not always be suitable for flicker-voltage measurements. Second, the three-phase power and current fluctuations occur simultaneously and the peak is about 2% times the minimum. This is interesting because it shows that the slip of induction type motors cannot prevent load fluctuations from showing up in the supply lines, unless the inertia of the load is high or the rate of power pulsation is high.

724

Lamp

Flicker

on Power

Systems

Chapter

22

Publication C 50-1943 American Standard Rotating Electrical Machinery of the American Standards Committee establishes the amount of pulsations for synchronous motors. Section 3-160 reads:
Pulsating Armature Current: When the driven load such as that of reciprocating type pumps, compressors, etc., requires a variable torque during each revolution, the combined installation shall have sufficient inertia in its rotating parts to limit the variations in motor armature current to a value not exceeding 66 percent of full load current. NOTE IThe basis of determining this variation shall be by oscillograph measurement and not by ammeter readings. A line shall be drawn on the oscillogram through the consecutive peaks of the current wave. This line is the envelope of the current wave. The variation is the difference between the maximum and minimum ordinates of this envelope. This variation shall not exceed 66 per cent of the maximum value of the rated full load current of the motor. (The maximum value of the motor armature current to be assumed as 1.41 times the rated full load current.) Adopted Standard 6-13-1923.

of load fluctuation, and the steady-state performance of moderately-sized induction motors as determined by test or circle diagram may be used in calculating flicker due to cyclic load variation of power factor with load, but specific data should be used where obtainable.

The variation of power factor of a synchronous motor during cyclic load fluctuations is a more complicated phenomenon. The average power factor is, of course, greatly
influenced by the supply voltage and by the field excita-

The above excerpt provides a basis for standardization and gives a criterion for a design unlikely to cause flicker. However, there are still possibilities that this amount of pulsation may at times result in flicker, particularly if the rate is between 6 and 12 cycles per second, and the supply line impedance is high. An analysis of Fig. 7 shows that with an induction motor both the current and power factor pulsate when the motor load varies, the power factor being highest when the load is highest as shown in the tabulation below. Usually, the armature time constant is high compared with the rate

tion. The variations from this average power factor due to load fluctuation is largely dependent upon the rate of the fluctuations as compared with the time constant of the field. For example, if the field time constant is 1 second and the load fluctuates once every 2 seconds the synchronous reactance of the machine determines the extent of the change in power factor. If, however, the power fluctuations are, say, 8 cycles per second, the transient reactance largely determines the change in power factor because the load swings are too rapid to demagnetize the field. Since in flicker problems, the change in load is of greater concern than the magnitude of the load, the average power factor is of no particular interest. The preferable proce-

Fig. 7-Oscillogram of current I, voltage E, and three phase power W of a 100-hp wound-rotor induction motor driving an air compressor. The voltage change which cannot be measured from the oscillogram caused objectionable flicker.

Fig. 8Vector diagrams illustrating method of obtaining magnitude and phase position of synchronous motor current and magnitude of bus voltage with change of load. X, is system reactance and X, is motor reactance.

Chapter 22

Lamp

Flicker

on Power

Systems

725

dure, if complete motor data are available, is to calculate the changes in the bus supply voltage to the motor due to changes in the load on the motor. The method is illustrated in the vector diagrams on Fig. 8. Vector diagram (a) shows the vector relations for a synchronous motor operating at full load and 80 percent power factor lead. E,, Ebue and E, are respectively the system voltage, bus supply voltage to the motor and the internal voltage of the motor. IR, and IX, are the voltage drops through the system impedance. IX, is the drop through the motor where X, may be the synchronous, transient or subtransient reactance depending upon the rate of load fluctuation compared to the time constant of the machine. Using diagram (a) as the starting point where the motor power factor angle 4 is known along with the average load, Ebue and all of the reactances, the change in bus voltage can be obtained as shown in vector diagram (b). For all sudden changes in load the system voltage, E,, and the internal voltage of the motor, E,, remain substantially constant. To determine the sudden dip in bus voltage it is necessary to calculate a curve of bus voltage against motor load or motor load change. This requires for each point on the curve that a magnitude of current be assumed and the voltage drop through the system and motor determined. This will locate the internal voltage E, with respect to the system voltage E, (In Fig. 8 E, and also E, in the diagrams (a) and (b) have the same magnitude).

The position of the voltage drops will then determine the position of the current vector as well as the bus voltage vector Ebus. Using the current, voltage (EbuB) and the angle between them the power can be found. With the curve of bus voltage against motor load change the voltage for any desired change in motor load can be obtained. The variation in reactive kva with real power is shown in Fig. 9 for a typical synchronous motor. These data are

Fig. 10Characteristics

of a typical induction

motor.

Fig. 9Characteristics of a typical synchronous motor at normal rated voltage. Curve A is for rapid changes in load from initial value and curve B is for slow changes.

for a power factor of 80 percent at full load, but for ordinary purposes the variations in reactive factor may be superimposed on the initial reactive factor. Curve A is for a rapid rate of fluctuation starting from full load 80 percent power factor; Curves B are for a rate slow compared to the field time constant with fixed terminal voltage. Motor Driven Intermittent Loads-In this category fall motor drives where the nature of the work calls for heavy overloads, and for cyclic loads of long and irregular period. Saw mills and coal cutters are typical examples of applications where heavy overloads, sometimes to the stalling point, are common and difficult to prevent. The motor currents in such installations vary rapidly from light load, through pull-out at heavy current and high power factor, to the high locked-rotor current at low power factor. Punch presses and shears are examples of applications where the load goes through wide variations, but where flywheels and other design features limit both the rate of application and magnitude of the load swings. Motors used to drive intermittent loads are likely to have been designed with special characteristics. If possible, the fluctuation in current and power factor should be obtained by test or from the manufacturer. In the absence of such specific data, Curve B of Fig. 9 may be used for slow cycling intermittent loads, and the curve of Fig. 10 may be used for applications where pull-out and stalling occur. Electric Furnaces-There are three general types of electric furnaces-resistance, induction, and arc. The resistance furnace usually causes no more flicker than any other resistance load of comparable size. Most induction

726

Lamp

Flicker

on Power

Systems

Chapter

22

ing the regulator settings or by a combination of several of these procedures. Forcing the furnace in this manner increases both the magnitude and the violence of the load swings. The type of the scrap being melted also affects the extent of the load swings, heavy scrap causing wider fluctuations than light scrap. The oscillogram of Fig. 12 represents a short part of a melting-down period of a 10 000-kva arc furnace, At times,

Fig. 11Three-phase furnaces operate at

melting arc furnace of the Heroult high frequency, and therefore,

type. are Fig. 12Oscillogram at start of heat in a 10000 kva Heroult A single-phase arc struck and type three-phase arc furnace. restruck 10 times in the space of 15 seconds before all three phases struck. After this initial period, all three phases struck and restruck 10 times with currents in all three phases fairly well balanced before the arcs became generally stable. A portion of this performance is shown on this figurelo.

connected to the power line through a frequency changer and consequently represent a fairly steady load. Three-phase steel melting arc furnaces of the Heroult type, illustrated in Fig. 11, are being used to a considerable extent to make high grade alloy steel, and frequently cause voltage flicker. While the average load factor and power factor of electric arc furnaces are as good or better than many other industrial devices, the problem of supplying them with power is usually much more difficult. During the melting down period, pieces of steel scrap will at times, more or less, completely bridge the electrodes, approximating a short circuit on the secondary side of the furnace transformer. Consequently, the melting down period is characterized by violent fluctuations of current at low power factors, single-phase. When the refining period is reached, the steel has been melted down to a pool and arc lengths can be maintained uniform by automatic electrode regulators, so that stable arcs can be held on all three electrodes. The refining period is, therefore, characterized by a steady three-phase load of high power factor. The size of load fluctuations during the melting down period is influenced by a number of factors, of which the rate of melting is perhaps the most important. The furnace-supply transformers have winding taps for control of the arc voltage and in the smaller sizes (about 6000 kva and below) have separate built-in reactors to limit the current and stabilize the arc. The rate of melting is subject to further control by means of electrode regulators. Sometimes the production of the furnaces is stepped up by raising the arc voltage, reducing the series reactance, rais-

the current variations occur at a periodicity approximating the rate of the most objectionable flicker. A graphic chart illustrating the variation of load over a longer period of operation is shown in Fig. 13. These two figures are reprints of figures from reference 10. Calculated curves in Fig. 14 show the electrical characteristics of a 10000-kva, three-phase arc furnace. These curves were prepared on the assumption that the maximum attainable current would be approximately twice normal at 50 percent power factor. The effective impedance of the arc (based on 11 500 volts in the primary) is plotted as the abscissa. For convenience, zero ohms, as plotted, represents the minimum arc resistance as determined by the so-called short circuit condition. Actually, at this point there is appreciable voltage drop at the electrode tips, and considerable arc energy; the curves are plotted in this manner only to show the working range. It is of interest that the point of maximum power is not that of maximum kva. The usual melt-down range is probably between the points corresponding to 0 and 10 ohms, the arcs fluctuating during this period so that the heating effect is some sort of an average between these limits. The refining range is probably above 10 ohms. It is difficult to obtain definite figures on the values of instantaneous swings in current and power factor for use

Chapter

22

Lamp

Flicker

on Power

Systems

Fig. 14---Electrical characteristics of a 10 000 kva, three-phase

arc furnace.

These values will give approximately the same flicker as the single-phase swings given in references 14 and 15. The curve values are not-the maximum possible swings for a given furnace size but are good values to use in estimating flicker. The frequency of occurrence of these swings corresponds to the Extremely Frequent classification as given

Fig. 15 Equivalent

kva swings in an electric arc furnace.

Fig. 13-Graphic charts at time of same heat shown on oscillogram of Fig. 12. Furnace swings occur approximately once a secondlO.

in flicker determinations, because an oscillograph must be used and the maximum swings cannot always be caught. On small furnaces, the current may reach a maximum of 3% times that at full load, but the process of reaching this value is usually through a series of small increments, and as noted previously the annoyance to lighting customers is largely a matter of the rate of change rather than the total change. The kva swings given in Figure 15 are equivalent swings.

in Table 1. Load swings can occur more rapidly, but their magnitudes are less than those in Fig. 15. These curves can be used in conjunction with the method suggested in Sec. 5, to estimate the amount of flicker. The information shown in Fig. 15, together with suitable system constants should give a fair approximation of the flicker voltage to be expected. Electric Welders-This is a class of equipment of great importance in power system flicker. Most welders have a smaller ``on time than off time, and consequently, the total energy consumed is small compared with the instantaneous demand. Fortunately, most welders are located in factories, where other processes require a large amount of power, and where the supply facilities are sufficiently heavy, so that no flicker trouble is experienced. In isolated cases, but nonetheless important, the welder may be the major load in the area, and serious flicker may be imposed on distribution systems adequate for ordinary loads.

728

Lamp

Flicker

on Power

Systems

Chapter 22

The more common types of electric welders are: (1) (2) (3) (4) Flash welders Pressure butt welders Projection welders Resistance welders (a) Spot (b) Seam

In welders the source voltage, usually 230, 460 or 2300 volts is stepped down to a few volts to send high current through the parts to be welded. Practically all welders in service are single-phase, although experimental threephase welders show promise. With flash welders, one piece is held rigidly, and the other is held in quasi-contact with it, with voltage applied. An arc is formed, heating the metal to incandescence, and the movable piece is made to follow to maintain the arc. The heating of the metal is partly by the passage of current and partly by burning with the arc. After a sufficient temperature and heat penetration has been obtained, the pieces are forced together under great pressure. In some cases, the power is cut off before this `upset; in others, the power is left on. The current, drawn during the flashing period, is irregular because of the instability of the arc, so that the flicker effect is obnoxious more than if the current were steady at its maximum value. The average power factor during flashing may be as high as 60 percent. At upset, it is about 40 percent. The flashing may last up to 20 or 30 seconds, but 10 seconds is more common. The duration of power during upset is usually short; of the order of 1/2 second. This type of welder may draw up to 1000 kva during flashing and about twice this loading at upset. Pressure butt welders are similar to flash welders, except for the important difference that the parts being welded are kept continuously in contact by a following pressure. The heating is produced primarily by contact resistance. From a power supply standpoint the butt welder is more desirable than the flash welder because the welding current once applied, is practically steady and the only flicker produced is at the time power is applied and removed. The range of currents and power factors is about that for flash welders. Projection welders are similar to pressure butt welders except that the latter usually join pieces of about equal size, and projection welders usually join small pieces to large ones. The current demand is usually smaller, but the operations are likely to be more frequent. In resistance welders current is applied through electrodes to the parts to be welded, usually thin sheets of steel or aluminum. The weld is accurately timed to bring the metal just to the welding temperature. The pieces are fused together in a small spot. In the spot welder, one or a few such spots completes the weld. In a seam welder, a long succession of spots produces the equivalent of a single continuous weld or seam. Resistance welders are characterized by large short-time currents. In spot welders, the current may be applied for only a few cycles (on a 60-cycle basis), with welds following one another in a fraction of a second up to about a minute. Thus, from a flicker standpoint there are a succession of individual volt-

Fig. 16Ignitron

timer for resistance welder.

age dips occurring at objectionably frequent intervals. Seam welders have an on duration of a few cycles followed by an off duration also of only a few cycles. The

Chapter 22

Lamp

Flicker

on Power

Systems

729

Fig. 17Typical

resistance welders---(a) welder.

spot welder, (b) seam

process is a continuous one while a given piece is in the machine, and since the periodicity of the welds is uniform, the flicker can be annoying even for relatively small voltage dips. The essence of good spot and seam welding is accurate control of the heat, consequently accurate magnitude and duration of current are necessary. Vacuum tubes are being used to a large extent for welder control functions because there are no wearing parts, and close and consistent regulation of the heat is possible. Fig. 16 shows a photograph of an ignitron electric timer and Fig. 17 shows a typical resistance welder.

Resistance welders drawing energy from all three phases greatly minimize flicker. Electronic devices are used to convert from the GO-cycle, 3-phase source to a single-phase output of lower frequency, say 10 cycles per second. On small welders, the stored energy of capacitors or inductors can often be used to minimize the peak demand from the source. MiscellaneousUnder this category come special equipment as electric shovels, heavy rolling mills, and similar installat ions. Most of these must be considered individually as to special features and power supply. Strip mining shovels frequently cause severe voltage dips in power systems, principally because of their large size and wide variation of their loads. The fast rate of load application is usually injurious to the power system principally by creating a wide band of voltage fluctuation, rather than flicker as it is commonly encountered. The site of mining operation is often at out-of-the-way locations where the power requirements for general purposes are small and hence, the normal power facilities are of low capacity, and very susceptible to flicker due to load changes. The large continuous rolling mills now used extensively in producing wide metal strip have imposed a new problem on the power industry. Like the electric shovel, these loads do not necessarily produce flicker in the customary sense of the word. The power supply is usually through motorgenerator sets without added flywheel effect. The load comes on and drops off in steps as the metal enters or leaves the rolls. The individual increments are not in themselves abrupt, a fraction of a second to over a second being required for the metal to enter a roll completely. A large hot strip mill and a typical load chart are shown in Figs. 18 and 19. The power drawn by a large continuous mill may build up to 30 000 kw in a period of 8 seconds, stay nearly con-

730

Lamp

Flicker

on Power

Systems

Chapter 22

A heavy cycling load of this kind may produce wide frequency variations on an isolated power supply system and wide load swings on an interconnected system. A power plant recording chart in Fig. 20 shows the power flow between the steel mill power plant and a large power pool. Fig. 21 (a) shows the hot strip mill load cycle and

Fig. 21(a)

Hot strip mill load. (b) Effect on frequency large interconnected system.

of

Fig. 21 (b) the results of calculations on how power surges of this kind cause frequency disturbances which travel as waves between the local power company to which the steel plant is connected and a larger power pool. III. LOCATION
Fig. 19Load chart for a hot strip rolling mill.

OF FLICKER

VOLTAGES in one

Load equipment may create flicker conditions or more of the following locations: (1) (2) (3) (4) Secondary distribution Primary lines Substation busses Generating stations

stant for a minute, and then drop to almost zero in another 8-second period, There may then be an off period of a minute followed by a repetition of the load cycle. The power source is usually ample so that no flicker is perceptible to the eye, but there is nevertheless a tendency for the voltage to weave up and down. This is undesirable because it widens the band of voltage regulation and may cause excessive operation of feeder voltage regulators. Automatic control of the excitation to the motor-generator sets to conform to the load variations is effective in minimizing these voltage swings.

Fig. 20Power flow between a steel mill and a large interconnected power system.

Any flicker in bus voltage of the generating station can be expected to show up at practically all points served by that station. Similarly if a substation bus flickers, all of the radial loads from that substation are affected. Primary line flicker affects all customers remote from the source of flicker, and to a lesser extent, some of those nearer the source of supply. Secondary circuit flicker is usually confined to an area immediately adjacent to the source of the disturbances. The location of flicker voltage, or the extent of the afflicted area, has a considerable influence on possible remedies. If the generating station busses are affected, there are usually no commercially practical means of remedying the situation on the power system, and the correction must usually be made at the utilization point. If a substation is affected, but the generation stations are not, then more tie lines or transmission at higher voltage can be employed, or a separate line run from the generating station to the affected area. Sometimes the utilization equipment itself can be corrected. If a primary line is affected, improvements can be made in either the power system or the utilization equipment. If the distribution-system alone is affected the correction may be made either on the system or the utilization device. If the utilization device is standard equipment, it is usually best to correct the distribution

Chapter

22

Lamp

Flicker

on Power Systems

731

system, and thus improve other loads as well. If the utilization device is special, it is probably more efficient to correct the device.

suitable for the supply of either single-phase or polyphase loads. When a synchronous motor draws additional power from the line it drops back in phase position. This causes a temporary drop in speed, but the flywheel effect of the IV. REMEDIAL MEASURES rotor tends to oppose this change and to give up temporarily part of its rotational energy. This results in a A large variety of corrective equipment and procedures cushioning of the rate of application of load to the power can be used to minimize flicker. Those most commonly system, and a material reduction in peak demand can be considered are : effected for loads of short durations as compared with one1. Motor gene:rator sets half of the natural period of electro-mechanical oscillation 2. Phase converters (see Chapter 13). The natural period usually ranges be3. Synchronous con dens tween $$ and 1 second, so that for loads lasting about ?& 4. Series capacitors second and less, substantial reductions in peak demand 5. Shunt capacitors m-g can be expected. Thus, synchronous-motor-driven 6. Voltage regulatoj l-s sets are quite suitable for spot and seam welders having 7. Booster transformners an on time of 1 to 10 cycles (60-cycle basis). Similarly, 8. Motor starters sudden increases or decreases of load are shielded from the 9. Excitation contra21 power system if the load factor is high, but the load is 10. Load control subject to short violent irregularities. This is true of elec11. Flywheels tric furnaces, for example, where the overall load factor is cc---L--- _I -- --12. i3ysr;em crlanges good, but there is considerable ``choppiness, sudden power factor changes and short-circuiting of individual phases. 3. Motor Generator Sets For this type of load, synchronous motor drives are nearly A corrective scheme using m-g sets is illustrated in Fig. as effective from the flicker standpoint as squirrel-cage 22. In general, it is probably true that a motor-generator induction, and preferable for other reasons. When an induction motor draws added power from the line, it drops in speed. Its output, in the normal working range, is closely proportional to the slip, that is, to the difference between synchronous and actual speed. If load is suddenly applied to a generator driven by a squirrel-cage induction-motor, the system does not feel the full effect until the motor-generator set has slowed down from nearly synchronous speed to full-load speed. In the meanwhile, the inertia of the rotating parts supplies the energy, and thus the rate at which power is drawn from the system is materially reduced. Furthermore, as in the case of synFig. 22-Motor-generator set. chronous driving motors, if the generator load consists of a series of short pulses, the load is off before its full effect set between the utilization device and the power system is transmitted to the power system, and the peak load on gives the maximum possible reduction in flicker, because the system is thereby decreased. Because an induction it is effective in minimizing three of the most undesirable motor must actually slow down, whereas a synchronous load characteristics: single phase, low power factor, and motor merely shifts in phase, the rate of load application sudden application Since the only tie between the motor to the power system is less for the induction than for the and the generator is the shaft, the disturbances due to synchronous motor. On an average, it takes an induction single-phase load or to low power factor are not transferred motor-generator set about one second to transfer full load to the power system. The reactance of the driving motor, to the source. In Fig. 3, it is shown that this delay alone in conjunction with the flywheel effect of the motor and results in doubling the threshold of flicker perception, as generator delay the transfer of a change in load to the compared with the perception due to sudden voltage dip power system. The rate at which the voltage drops is of equal magnitude. therefore lessened and the eye is less likely to perceive If the load pulses last several seconds, the power drawn this flicker. from the system levels off to the amount of generator load The motor-generator set is probably the costliest arplus losses for either a synchronous or squirrel-cage motor rangement, heaviest, least efficient, and occupies more drive. The voltage drop in the power system during this floor space than any of the various corrective devices that steady load period is usually about the same for either can be used. But the m-g set has the advantage of conthe induction or synchronous motor drive, assuming that sisting entirely of standard equipment, and is, therefore, the excitation of the synchronous motor is constant. By reliable and well understood apparatus. The motor end increasing the synchronous-motor excitation with the load may be synchronous, squirrel-cage induction, or wound the final regulation of the system can be made very small. rotor induction, the latter usually being provided with a However, from the standpoint of flicker such excitation flywheel and slip regulator. The generator end may be changes are usually imperceptible because of the time

732

Lamp

Flicker

on Power

Systems

Chapter

22

required for correction. Thus, from the flicker standpoint, the principal superiority of the induction motor to the synchronous motor is the doubling of the threshold of perception, because of slower load application. This is particularly so for short pulses of power, say x second and less, where the induction set draws considerably lower peaks than the synchronous set. A further material reduction in flicker can be effected by the use of motor generator sets equipped with flywheels. In such cases a wound-rotor-induction motor is used, and additional rotor or secondary resistance is connected externally. By this means, the full-load slip of the motor can be increased from 1 or 2 percent to 10 percent or more. In order to transfer full load to the system, the set must then slow down considerably and the fullest advantage is thus taken of the inertia of the set and the additional flywheel. The extent to which improvement by this means may be carried is limited by cost and each case must be considered on its own merits. Limitation of peak demand is probably not feasible for loads in excess of about 3 seconds, but the reduction of rate of load application may nevertheless be of benefit. Figures 23 and 24 bring out in graphic form the points discussed above. These curves were calculated using typical machine constants, and to facilitate computation, losses were neglected except when used to calculate speed changes on the induction sets. The curves of Fig. 23 are for a load on 1% seconds and off 49; seconds. Curve (a) represents the load drawn by the synchronous set, and shows that it takes approximately 0.2 seconds for the system load to equal the generator load, and also that an overswing of about 35 percent makes up for the deficiency between input and output during the first 0.2 second. A similar swing occurs when the load is dropped. Curve (b) shows the load drawn by a standard squirrel cage induction motor subject to the same load

600 400 200

600

200 0 TIME-SECONDS

Fig. 24Curves showing the relation between the power supplied by the generator and the power taken from the system for motor-generator sets using three types of motors. Generator load on for 0.1 second.

Fig. 23Curves showing the relation between the power supplied by the generator and the power taken from the system for motor-generator sets using three types of motors. Generator load on for 1.5 seconds.

cycle. It can be seen that the system load builds up at about half the rate as for the synchronous motor, and that it does not become equal to the applied load until the end of the load application. The system load never exceeds the applied load disregarding, of course, m-g set losses, and the difference between input and output during the early part of the load cycle is compensated by a similar exponential continuance of load on the system for some time after the applied load has ceased. Curve (c) is for a wound rotor motor with a constant secondary resistance and a flywheel. The relation between slip, flywheel effect and load cycle is such that although the generator load goes on and off, the system load never drops to zero. The rate of load application is very low, and the system peak is only about a third of the load peak. The curves of Fig. 24 are for a load cycle of 0.1 second on and 5.9 seconds off. Curve (a) is for a synchronous motor and shows that the peak system load is about two-thirds of the generator load. Curve (b) is for the squirrel cage set and shows a system peak of less than of the generator peak. Curve (c) is for the flywheel set and shows a system peak of about 3 percent of the load peak. Figures 23 and 24 are of interest in illustrating the manner in which motor-generator sets transfer power from load to line, and suggest the conditions under which the various motors are most suitable. As pointed out previously, the phase balancing and power factor improvement qualities are usually the most valuable factors in the correction of flicker. There are so many variables in load, power factor, duty cycle, etc., that general figures on the improvement that can be expected may be open to criticism. For very approximate purposes, however, it can be expected that if the load changes last one second or more, either synchronous or squirrel cage induction sets without flywheels reduce the voltage drop to ?& for single-phase loads and to g for polyphase loads. The perceptibility of the flicker is re-

Chapter

22

Lamp

Flicker

on Power

Systems

733

duced still further by the slower rate at which the voltage dips, particularly with the induction set. For loads of very short duration such as ?& second and less the voltage drop may be reduced to l/10 or even l/20. Motor generator sets may be had with either single- or three-phase generators. Even when the generator is single phase, it is customary to use a three-phase star stator winding using only two legs in series. The third phase is wound for possible future use, or to increase synchronizing power if paralleled with other units, or dummy coils may be placed in the slots. If single-phase loads are to be carried, the field must be built with low resistance damper windings to minimize rotor heating. In the larger sizes, single-phase machines are mounted on springs to minimize vibration due to the pulsating. torque caused by singlephase operation. When more than one utilization device causing flicker is involved, the question of a single m-g set versus an m-g set for each such load must be answered. In these cases it is very important to consider the regulation of the generator of the set and how constant a voltage is required by the utilization devices. For example, it frequently happens that a factory is using several electric welders which produce 5 percent voltage dips of very objectionable frequency. This 5 percent drop usually does not affect the performance of the welders, and they could be operated at random on the power system. If a motor-generator set is to be used, however, the transient reactance of the generator is apt to be as high as 35 percent based on its rated current, and, assuming that the welder reactive current equals the generator rating, a 35 percent, drop in voltage would occur. If only one welder is operated at a time, this is quite satisfactory, as the welder tap can be set on the basis of closed circuit voltage, that is, the regulation of the generator can be taken into account. If, however, another welder is operated simultaneously, even though on another phase, the additional voltage drop, uncompensated by the welder tap, is enough to spoil the weld. In order to operate several choppy loads simultaneously from the same m-g set, it is therefore necessary to use an oversize generator (from a thermal standpoint) to keep the regulation within required limits. Alternate solutions are to interlock utilization devices so that they cannot operate simultaneously or to provide separate m-g sets for each device. Another alternative is to use one common driving motor and several separate generators on the same shaft. The separate m-g set plan has the advantage of permitting operation at partial capacity in case of damage to one set, but is costlier.

vices such as inductances and capacitors, or in rotating equipment with mechanical inertia. Except for small sizes, the static equipment has not yet been found commercially practical. A lack of appreciation of this fundamental energy requirement has led to frequent proposals of schemes attempting single-phase to polyphase conversion by transformer connection. Fig. 25 is typical of these schemes. It

Fig. 25Unsound attempt to supply balanced power to a single-phase load.

three-phase

is not only completely ineffective for its intended purpose, but is also wasteful of transformer capacity. Although the transformers are all loaded equally, the currents drawn from the source as shown by the current arrows, are still single-phase, and a single-phase transformer is, therefore, preferable. The most familiar type of phase converter is that shown in Fig. 26. It has been extensively used in railway electrifications to convert single-phase power from the contact system to three-phase power for the locomotive motors; this is merely the converse of the phase-balance. As shown, a rotating two-phase machine is connected to the three-

4. Phase Balancers
In industrial plants a large percentage of the potential causes of flicker are single-phase devices. A discussion of phase balancers is, therefore, of interest, although there have been few commercially installed. In a single-phase circuit the flow of power pulsates at a frequency twice that of the alternating supply, whereas in a balanced polyphase circuit the flow of power is uniform. Therefore, in order to effect a conversion between a single-phase and a polyphase system, some energy storage is necessary. This storage may be made in static de-

Fig. 26Schematic diagram for phase converter used extensively on railway electrificationa to convert single-phase power from the trolley to three-phase power for the locomotive motors. A rotating two-phase machine is connected through the equivalent of a Scott-connected transformer to the three-phase power system.

734

Lamp

Flicker

on Power

Systems

Chapter 22

phase power system through the equivalent of a Scottconnected transformer, which also serves as the primary for the single-phase load winding. The two-phase machine may be of the induction type and act as a phase converter only, or it may be synchronous and used for power factor correction as well. Because of the regulation of the machine, the source currents are not balanced during variableload conditions, unless the taps on the transformer winding are varied. From this point of view, it is not very suitable for choppy loads. Where there are several separate single-phase loads to be served, the capacity of a converter of this type must be equal to the sum of the individual loads. The series type of phase converter is shown in Fig. 27. This is probably most efficient for conversion from three-

Fig. 27Series type of phase converter single phase.

from three phase to

phase to single-phase, where the single-phase load is not expected to grow, cannot be distributed between phases, and where no power factor correction is required. It consists of a counter-rotational induction-type series machine, connected through transformers in such a manner as to offer a high impedance to negative-sequence current between the single-phase load and the three-phase supply.

When a single-phase load is suddenly applied, a magnetizing transient results, so that part of the negative-sequence component of load current is passed on the source. Although this transient subsides in about 0.1 second, it detracts considerably from the value of the scheme for use with choppy loads. The series impedance balancer shown in Fig. 28 consists of an auxiliary induction-type machine in series with the polyphase supply and with the main shunt machine. The single-phase load is drawn from between the two. The series machine rotates oppositely to normal direction for positive-sequence applied voltage, and therefore, offers high impedance to negative-sequence currents and low impedance to positive-sequence currents. The shunt machine therefore takes the negative-sequence component of load current. The positive-sequence component of load current is taken by the system if the shunt is an induction type unit. If a synchronous type unit is used for the shunt machine, it can also take the wattless component of load current with suitable control of excitation. As with the series phase converter, the series machine does not immediately respond to load changes, and temporarily (for about 0.1 second) some unbalanced current is drawn from the source. The scheme, like the series phase balancer, is inherent in its action, no regulators being required unless power factor correction is used. This method has one important advantage over the previous two schemes in that the size of the shunt machine need only be enough to take care of the maximum unbalance of load. For example, if there are a number of individual single-phase loads as illustrated in Fig. 29, they may be distributed

Fig. 29Effective use of a synchronous condenser in connection with a fluctuating load.

between the phases, and the shunt machine need carry only the unbalance component. The series machine must, however, have enough capacity to carry the total positive sequence current. Phase balancers, as a class, are not particularly suitable for flicker elimination except perhaps in borderline cases where only a moderate improvement (perhaps a one-half reduction in voltage dip) is required. In this case they may be the cheapest and most efficient remedy.

5. Synchronous

Condensers

Fig. 28Series impedance

type of phase balancer.

The voltage dip on a power system resulting from a suddenly applied load is equal to the vector product of the current and the system impedance giving proper consideration to vector positions. Consequently, one way of reducing flicker is to reduce the system impedance. Usu-

Chapter

22

Lamp

Flicker

on Power

Systems

735

ally, the system impedance is predominantly inductive, and flicker is caused by current of low power factor so that most of the voltage drop is due to the reactive component of the system impedance. For example, suppose that the system impedance based on the load current is 1 percent resistance and 4 percent reactance and the load is at 50 percent power factor. A close approximation of voltage drop may be obtained by adding only those components of impedance drop that are in phase with the voltage. Thus, the resistance component of line drop is the 1 percent resistance times the 0.5 unit of current or 45 percent, and the reactive component of line drop is the 4 percent reactance times the 0.866 unit of current (for 50 percent power factor) or 3.5 percent. The total voltage drop is therefore 4 percent, of which 3.5 percent is due to system reactance. This predominance of reactive component has led to frequent proposals to use synchronous condensers in parallel with the system as a means of reducing system reactance and thus improving flicker conditions. This method, while feasible in principle, is not usually economical in practice, as a brief consideration shows. The system reactance to a customers service point may range from a fraction of a percent to 10 or more, but on an average is probably around 5 percent, based on the customers kva demand. The subtransient reactance of a standard synchronous condenser is around 25 percent of its rating. Therefore, if a synchronous condenser of the same kva rating as the load is installed, the resultant 5x25 reactance is - 3. = 4.2 percent and the flicker voltage is reduced to only 4g=84 percent of its value without the . condenser. The effectiveness of a synchronous condenser can be much improved by the use of reactors between the power system and the load and operating the condenser from the load bus, as shown by Fig. 29. This scheme permits greater voltage fluctuations on the condenser and, therefore, causes it to bear a greater proportion of the fluctuating component of current. The customers bus voltage, of course, undergoes the same voltage fluctuation, and this fact plus the fact that only a limited amount of series reactance can be used without unstable condenser operation, limits the extent of improvement. In most instances, it is likely that a reduction of flicker to one-half its uncompensated value is the economic limit of correction by this means. Where only this amount of correction is sufficient, the synchronous condenser and series reactor scheme may be the best economic solution, considering the power factor correction and control of voltage level afforded by the machine. The suggestion has been made of using a driving motor for the synchronous condenser to permit higher values of series reactance without instability. This arrangement is the equivalent of a motor-generator set with a reactor paralleling the motor and generator ends. This scheme has never been used in practice, but calculations of performance and cost estimates indicate that there is little advantage compared with the straight m-g set or condenser-reactor schemes. The benefits from the use of synchronous condensers

for flicker reduction depends in a large measure upon how low the subtransient and transient reactances can be made. The modern standard low-speed salient-pole synchronous condenser has been developed primarily for power factor correction and voltage control, and low-cost and low-loss condensers have relatively high reactance. A typical machine has subtransient and negative sequence reactances of about 25 percent and a transient reactance of 35 percent. A reduction in these reactances usually results in both higher costs and losses. The high-speed (3600 rpm) cylindrical-rotor type of machine inherently has lower reactantes, perhaps one-half or less, but the cost and losses are both greater. In larger sizes and where other circumstances are favorable, the overall economy may justify the use of outdoor highspeed hydrogen-cooled synchronous condensers of low reactance. Another way to decrease the reactance of the synchronous condenser is to use capacitors in series with the machine leads. The capacitive reactance partially nullifies the machines inductive reactance giving a lower net reactance. This scheme theoretically should be quite effective and economical. However, the series capacitors may cause the synchronous condenser to hunt. The boundaries of satisfactory operation have not been fully explored, and predetermination is difficult. It is expected that after an experimental installation of this form of compensation is made that practical information will be available.

6. Series Capacitors
A general treatment of the use of capacitors in power systems is given in Chapter 8. The following discussion is concerned primarily with those aspects of capacitor application that are related to the problem of lamp flicker. There are two main uses of series capacitors, depending whether they correct for the inductance of the supply or for that of the load. Their most familiar use is for line drop compensation; the application to equipment correction is more recent and shows much promise, as it improves conditions in the entire system, whereas the line capacitors benefit only those customers beyond the point of capacitor installation. Being in series with the entire power circuit, series capacitors are instantaneous in their corrective effect. This is perhaps their most valuable advantage because any change in line current causes an immediate change in compensating voltage. Another advantage is that they generate lagging reactive kva proportional to the square of the current, thereby improving the power factor. Series Capacitors Connected in LineFig. 30 shows in (a) a layout ordinarily favorable to the application of series capacitors. The transmission substation is assumed to have bus voltage regulation so that the voltage is fairly constant. The step-down transformer bank and the lowvoltage line feed a distribution substation serving the fluctuating load and lighting loads; no loads are served at intermediate points between the substations. The series capacitor may be installed near the transmission substation, as shown in (b), or near the distribution substation. Another alternative is to install the capacitors between the transmission substation bus and the step-down transformer (depending upon which voltage is more suitable

736

Lamp

Flicker

on Power

Systems

Chapter

22

The voltage along the line is for standard capacitors). shown by the diagram at (c), Curve A showing the uncompensated voltage and B the compensated voltage. The point of interest emphasized by (c) is that the compensating voltage is introduced in one step while the voltage drop along the line is uniform. For this simple case with

Fig. 3@---Typical application

of series capacitors.

(a) Layout ordinarily favorable to application of series capacitors (b) Location of series capacitor (c) (A) Without capacitors; (B) with capacitors.

no intermediate line loads, the voltage gradient along the line is unimportant, and, subject to limitations outlined later, complete voltage-drop compensation at the distribution substation may be secured. The vector diagrams for series capacitors at various power factors are shown in Fig. 31. These diagrams show that only the inductive component of line impedance is compensated by the capacitor. However, if the power factor of the load increment is low and constant, it is possible to over-compensate for the system reactance, and thus partly or completely nullify the resistance component of line drop. With variable loads and power factors this procedure can cause undesirable voltage-regulation characteristics and therefore each case of over-compensation must be considered on its own merits. Where there are distributed loads along a line, it is necessary to consider the location of the capacitors. The capacitor gives its full voltage boost at the point of its installation, and therefore loads immediately ahead and behind the capacitor differ in voltage by the amount of boost in the capacitor. In general, the best capacitor locaFig. 31The vector diagrams show the voltage drop across the series capacitor required if a capacitor is added so that the sending voltage will be the same as the load center voltage when the load power factor is (a) 90 percent; (b) 75 percent; (c) 60 percent.

Chapter 22

Lamp

Flicker

on Power

Systems

737

tion is one-third the electrical distance between the source and the flicker-producing load, as shown by Fig. 32. In principle series capacitors are effective in reducing flicker caused by practically all types of fluctuating loads. However, their effect is only beyond their point of instal-

Fig. 34Series capacitor installed with a welding load to reduce kilovolt ampere demand and improve power factor.

Fig. 32Percent voltage regulation-in general, by placing the series capacitor about l/3of the electrical distance between the source and the load, the voltage on both sides of it are kept within plus or minus limits in which flicker is not objectionable.

Capacitors in Series with the Equipment-This application is limited to utilization equipment with a constant inductive reactance, for which it is possible to compensate with a series capacitor, so that the load drawn from the supply circuit is practically at unity power factor at all times. Thus, although the power drawn from the line is still fluctuating, the resultant flicker voltage is greatly reduced. Figure 34 shows such compensation applied to a welding transformer. Inasmuch as the load itself is corrected, the benefits are felt all over the supply system. Several such applications have been successfully made to spot and seam welders (see reference 3). 7. Shunt Capacitors
Contrary to frequent misconceptions, permanently connected shunt capacitors are of no benefit whatever in minimizing flicker; in fact, they may make it slightly worse. An example shows the reason readily. A system with 10 percent inductive reactance in the supply leads, serving an intermittent load having an inductive reactance of 100 percent is shown in Fig. 35 (a). Resistance in both line and load will be neglected to simplify the example, but the same general effect will be observed if resistance

lation; hence they do not correct the system as a whole. For example, a series capacitor installed just ahead of substation B in Fig. 33 may remove most of the voltage fluctuation on that bus. However, at Station A, there may still be considerable voltage fluctuation, as the series capacitors do not correct the supply circuits. Another point to be noted from Fig. 33 is that the series capacitor must be large enough to carry all loads beyond its point of installation. Consequently, if the flicker-producing load

Fig. 35Shunt Fig. 33Series capacitor must be large enough substation load. to carry total

capacitors are not effective in reducing voltage dips.

were present.

When the switch is open E,=Es.

When the

is small as compared with normal load, the cost of the series capacitor is too high for the correction obtained. Series capacitors are therefore economical primarily where the flicker load is a large portion of the total, where the circuit resistance is equal or lower than the reactance, where the flicker-producing load is of low power factor, and where the supply circuits are fairly long. Under certain circumstances series capacitors will produce, in conjunction with other apparatus, voltage or current surges in the line. The magnetizing inrush current of transformer banks, and the self-excitation of synchronous or induction motors are some of the factors causing this phenomenon, which is too involved for treatment here, but is discussed in items 4 and 5 of the table of references.

+jloo Es = 91 +j10+j100 percent Es. Fig. 35 (b) shows a similar circuit except a capacitor having a reactance equal and opposite to that of the load is permanently connected in the circuit. When switch is closed, the voltage at EL= the switch is open, the voltage EL= percent Es. pedance is
c-jlOO) (+jloo) = o.

-do0 Es= 111 +j10 -jlOO When the switch is closed, the net load im-j100+j100

This

means that

the

combination of the capacitor and reactor draws no current from the source, and E,=Es. Thus, comparing the two cases, without the capacitor the voltage drops from 100 percent to 91 percent, a change of 9 percent. With ca-

738

Lamp

Flicker

on Power

Systems

Chapter 22

pacitors, the voltage drops from 111 percent to 100 percent, a change of 11 percent. Shunt capacitors connected to utilization equipment so that they are switched in accordance with load, reduce voltage drop. To be effective, the utilization device must draw a current that is substantially constant in magnitude and power factor during the on period, as, for example, some forms of resistance welders on which long runs are made without change of set-up. Motor starting is one example of an application to which shunt capacitors cannot be used effectively in this manner for flicker reduction. Motor inrush current approximates six times full load. If this is neutralized by a shunt capacitor, the initial voltage dip is greatly reduced. However, when the motor comes up to speed, the voltage rises above the initial voltage.

8. Voltage Regulators
Voltage regulators are also totally unsuited to correcting flicker. This statement applies both to generator voltage regulators, or to step- or induction-type feeder regulators. These devices operate only when the voltage changes; furthermore there is a time lag before voltage is restored to normal. As shown in Fig. 3, abrupt changes in voltage, the ones that voltage regulators cannot eliminate, are the very ones to which the human eye is most sensitive. Consequently, the flicker is perceived before the regulator can even start. It is sometimes thought that an electronic regulator and exciter can eliminate this difficulty and prevent voltage dips. However, the field time constant of the generator which in large units is as high as 10 seconds and even in very small machines may be one second, makes correction by this means impossible.

completely. Satisfactory results can often be obtained by omitting the resistor, and in such cases, the apparatus becomes simply a transformer with an air gap in its magnetic circuit. Despite the technical simplicity of this scheme, it has practical and economic limitations. It is apparent that the improvement in the lighting circuit is obtained at the expense of the flicker-producing load. This limits the application to cases where the lighting load is only a small proportion of the total. In general, the equipment must be individually designed for a specific set of conditions, since the proportions and size are affected by the line voltage, line drop, total current, and ratio of loads. Should system changes necessitate its removal, there is small likelihood of being able to use the compensating transformer elsewhere. The cost of the apparatus is rather high because it is not standard.

10. Motor Starters

As pointed out under Utilization Equipment, most motors can be started directly across the line because even the larger sizes are usually supplied from heavy feeders compared to the size of the motor. Where this is not the case, a starter may be required if the starting is frequent. It is difficult to generalize on the question of motor starting, because individual cases vary with the motor size, type, and the starting torque of both motor and load. Starting compensators are now being used much less than formerly. This is due largely to the acceptance of across-the-line starting, but also to the realization that the two voltage dips caused by the compensator may be as objectionable as one larger dip when starting across the line. In this respect reactor starting is superior, because the circuit is not opened at transition, and the reactorshort-circuiting operation may not result in a noticeable voltage dip if the motor is substantially up to speed. A reactor starter causes a greater initial voltage drop than a compensator, because the starting kva is decreased only directly as the starting voltage and not as the square of the voltage. When the continuous-load rating of the feeder is the same as of the motor, the use of wound-rotor motors with stepped-resistance starters in the rotor circuits usually avoids annoying flicker. The cost of the motor and control is greater, but where the motor is near the end of a long line and is started frequently, this may be the most economical choice. Where motors are started infrequently, but where the resultant voltage dip is still objectionable, some form of increment starter may be warranted. In a starter of this type, the stator current is increased in steps until the motor rotates, and the remaining impedance is cut out of the circuit after the motor has reached full speed. There are no standard starters of this type on the market, and the few that have been built have been specially designed for the particular service. In general, they represent a combination of auto-transformer and reactor starting, the switching being done without opening the circuit during the entire sequence. Resistance starters in the stator circuits have been employed. On small integral horsepower motors the simplest

9. Compensating

Transformers

As illustrated in Fig. 36, a compensating transformer is similar in effect to a line drop compensator used in voltage regulator control except that the size of the elements is that of a power device rather than that of an instrument. The current drawn by the flicker-producing load passes through a resistance and reactance branch, and the voltage

Fig. 36Compensating transformer can be used in very special cases to reduce voltage dips.

drop thus created is added to the lighting-load voltage by means of a series transformer. By proper selection of the resistance, reactance, and series-transformer ratio, the flicker in the lighting circuit may be eliminated almost

Chapter

22

Lamp

Flicker

on Power .

Systems

739

and cheapest of these is a single-step resistor which is cut out after the motor comes up to speed. As with reactor starting used on larger motors, the short-circuiting of the resistor does not usually cause a noticeable voltage dip, and the initial dip of course is considerably reduced. Resistance starters should be adjustable for individual requirements; in extreme conditions a variable resistor may be desirable. These starters are in general more expensive and more difficult to maintain by unskilled attendants.

11. Excitation

Control

This involves single-step increments of the field excitation of synchronous motors by switches actuated by the equipment causing the flicker. This method is generally ineffective in eliminating flicker caused by abrupt voltage dips as explained under Voltage Regulators. However, it can reduce considerably the width of the band of voltage regulation, which annoys power-supply companies by causing too frequent operation of feeder-voltage regulators as they attempt to compensate for the voltage swings. Such swings are caused by continuous strip rolling mills, large electric shovels, etc., where the variations of load are large, but where the rates of application and removal are moderate, say 10 to 30 percent per second.

Fig. 37System layout. (a)

Fluctuating load on substation bus affected all loads fed from

load feeders separated from rest of the load.

bus. (b) Fluctuating

at a time when the lighting load is low. Control of load is not a very general solution to reduction of flicker, and it is employed in but few cases.

12. Load Control

In some cases it is possible to minimize lamp flicker by controlling manufacturing processes. For example, in a plant operating two or three resistance welders, it may be possible to provide interlocks so that not more than one is operated at the same instant. A remedy of this kind is only possible if the on time is short compared to the off time, otherwise the production rate would be slowed up considerably. Similarly in arc-furnace work the violence of the current swings during melting can be reduced by lowering production rate during this phase of the cycle. It is also possible to perform flicker-producing operations

13. Flywheels
A general discussion of the effect of flywheels is given under Motor-Generator Sets, but the same principles apply to direct-driven apparatus. This method has considerable value for mechanical loads having short durations with long off periods, such as shears, punch presses, etc.

14. System Changes

In practically all cases of flicker caused by utilization equipment, there is a direct relationship between the amount of the flicker and the size of the power supply system. For example, assume that a weldercausesa three-

740

Lamp

Flicker

on Power

Systems

Chapter 22 REFERENCES

percent voltage flicker on a residential substation, where only one percent is acceptable. Tripling the size of the supply to the substation would reduce the flicker to the required level, and this would constitute one way of eliminating the flicker. If this were done by multiplying the number of incoming lines and transformer banks by three it would probably be the most costly of all possible corrective measures. Usually more economical system changes can be made. A common form of substation supply with two or more feeders from the generating station paralleled to a single bus is shown in Fig. 37(a). With this arrangement, all loads fed from the substation are subjected to any flicker produced on the outgoing feeders. Figure 37 (b) shows a low voltage bus divided into two sections, one for residential and commercial loads, the other for industrial loads. This layout is based on the fact that voltage fluctuations objectionable to residential customers are acceptable to industrial users. There is probably a greater flicker tolerance in shop work than in residence lighting, and industrial plants are usually willing to accept flicker when it is caused by their own operation. Other methods of stiffening the power system involve changing the voltage of the supply line, tapping nearby high-voltage, high-capacity lines, adding more transformer capacity, or running a separate line to the flicker-producing load. Local conditions determine what remedial measures are most suitable in a particular case. Occasionally system increases are justified if the additional capacity may be needed later anyway.

15. Comparison Chart

A reference chart showing at a glance the remedial measures available and those most promising for a particular type of flicker is shown in Table 2. Inasmuch as the best technical solution may not be the most economical, the remedies are compared from both points of view.

1. The Visual Perception and Tolerance of Flicker, prepared by Utilities Coordinated Research, Inc.--New York, 1937. 2. Lamp Flicker Awaits Ideal Motor Starter, by L. W. Clark, Electrical World, April 9, 1938. Correction of Resistance-Welding Machines by 3. Power-Factor Series Capacitors, by L. G. Levoy, Jr., A.I.E.E. Transactions, 1940. 4. Analysis of Series Capacitor Application Problems, by Concordia and Butler, A.I.E.E. Transactions, 1937. Vol. 56. 5. Self-Excitation of Induction Motors with Series Capacitors, by C. F. Wagner, A.I.E.E. Paper No. 41-139. Presented at Summer Convention, Yellowstone Park. 6. A Lamp Flicker Slide-Rule, by C. P. Xenis and W. Perine, Presented at E.E.I. Transmission and Distribution Committee Meeting, Chicago, May 5, 1937. Machines, Committee 7. Power Supply for Resistance-Welding on Electric Welding, A.I.E.E. Transactions, 1940. Vol. 59. Machines-Factory 8. Power Supply for Resistance-Welding Wiring for Resistance Welders, Committee on Electric Welding, A.I.E.E. paper 41-82Contains a Number of Examples. 9. Power Supply for Welding, by A. S. Douglass and L. W. Clark, The American Welding Society Journal, October 1937. 10. Large Electric Arc Furnaces-Performance and Power Supply, by B. M. Jones and C. M. Stearns, A.I.E.E. Transactions, 1941. Vol. 60. 11. Arc Furnace Loads on Long Transmission Lines, by T. G. Le Clair, A.I.E.E. Transactions, 1940. Vol. 59. 12. 10 000 kva Series Capacitor Improves Voltage in 66 Kv. line Supplying Large Electric Furnace Load, B. M. Jones, J. M. Arthur, C. M. Stearns, A. A. Johnson, A.I.E.E. Transactions. Vol. 67, 1948. 13. Voltage Translator Scheme Cuts Light Flicker due to Welders, R. 0. Askey, Electrical World, January 6, 1945, page 63. 14. Electric Arc Furnaces and Equipment Producing Heavy Fluctuations, Part II-the solutions, by B. M. Jones. Presented before E.E.I. Electrical Equipment Committee, Old Point Comfort, Va., October 10, 1950. 15. Power Company Service to Arc Furnaces, by L. W. Clark, A.I.E.E. Transactions 1935.