See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/3169881 Harmonic Measurements in Industrial Power Systems Article in IEEE Transactions on Industry Applications · February 1995 DOI: 10.1109/28.363032 · Source: IEEE Xplore CITATIONS READS 36 1,618 1 author: P.E. Sutherland ABB 67 PUBLICATIONS 463 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Power Generation View project Harmonics View project All content following this page was uploaded by P.E. Sutherland on 07 December 2015. The user has requested enhancement of the downloaded file. IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 1, JANUARY/FEBRUARY 1995 175 Harmonic Measurements in Industrial Power Systems Peter E. Sutherland, Member, IEEE Abstract- Harmonic measurements are made in industrial power systems in order to: (a) aid in the design of capacitor or filter banks, (b) verify the design and installation of capacitor or filter banks, (c) verify compliancewith utility harmonic distortion requirements, and (d) investigate suspected harmonic problems. The results of these measurements are used in design calculations, verification, comparison with standards, and system modeling. Each of these objectives will affect the choice of a measurement approach. The selection of the measured quantities,measurement points in the system, and the types of instrumentsand transducers should be based upon the measurement objective. Once measurements are taken, additional calculations must be made to put the results into a useful form. The measurement results will then provide a firm basis for further engineering work. current injections from standard power conversion equipment, such as six- and twelve-pulse drives, may be estimated without measurement. However, the loading and diversity factors may not be known. In addition, the wide variety of harmonic producing equipment in a facility may make estimation difficult. The harmonic measurements should be taken at times when the largest harmonic sources are in steady-state operation at maximum loading. Harmonic measurements are also needed to determine the distortion present from the utility system. This factor can only be determined by measurement. The harmonic measurement data are then used to calculate the expected harmonic currents and voltages to be experienced by the proposed capacitor or filter bank. I. INTRODUCTION H ARMONIC analysis studies of industrial power systems require measurement data. An approximate calculation may be made using estimates of harmonic magnitudes derived from equipment nameplate data, but this might not reflect operating conditions. IEEE Standard 519-1992 [I] contains recommended practices for performing harmonic measurements. The harmonic measurement process begins with the choice of equipment and techniques that will determine the validity of the results. General purpose spectrum analysis instruments provide a wide variety of possible measurement modes, from which the proper settings for power system measurements must be chosen. Special-purpose power system harmonic analyzers exist which have fewer choices of measurement modes available, and thus require an understanding of when and how they can be used. Transducers that provide the low voltage input for instruments are a critical link in the measurement process. Measurement points in the system depend upon what analysis is to be done with the measurement results. 11. PURPOSES OF HARMONIC MEASUREMENT B. Harmonic Filter and Capacitor Bank Installation After the design process is completed, and a capacitor or filter bank has been installed, it should be checked to determine whether the desired results have been obtained and that the bank has been applied within its design limits. Power factor measurements and fundamental frequency voltage and current measurements provide initial confirmationof correct operation. Measurements of the harmonic voltages at, and currents into, the bank should be compared with the specified harmonic voltage and current capabilities of the bank. These measurements are a baseline for tracking the future performance of the device. Additional measurements should be taken when when any significant change is made in the power system. A bank that has been properly designed for one set of conditions may fail when subjected to conditions not envisioned at the time of installation. Harmonic current and voltage levels at a common bus can vary with not only the magnitude, but the phase angle of the harmonic sources. For this reason, a single measurement alone may not indicate worst case conditions. Harmonic measurements should always be evaluated in conjunction with harmonic calculations. A. Harmonic Filter Design and Capacitor Bank Application Before power factor correction capacitors are applied in a facility where significant sources of harmonic currents andor voltages are present, it is usually necessary to perform a harmonic analysis to determine whether a capacitor bank may be used, or whether a harmonic filter is needed. Harmonic Paper ICPSD 94-14, approved for publication by the Power Systems Protection Committee of the IEEE Industry Applications Society for presentation at the 1994 IEEEAAS Industrial & Commercial Power Systems Technical Conference, Irvine, CA, May 1-5. Manuscript released for publication July 20, 1994. The author is with GE Industrial & Power Systems, Installation and Service Engineering, General Electric Company, Albany, NY 12205 USA. IEEE Log Number 9406625. C. Compliance with Utility and IEEE 519 Requirements The requirements specified by an electric utility company for harmonic injection by industrial customers may be checked by measurements, but additional calculations are often needed. IEEE Standard 5 19 [ 11 specifies allowable levels of harmonic currents at the “point of common coupling.” Harmonic measurements, coupled with harmonic analysis of both the worst case conditions and of the effects of the addition of power factor correction capacitors or harmonic filters, will indicate whether or not a possible harmonic problem exists, and can also serve to verify the correctness of the solution. 0093-9994/95$04.00 0 1995 IEEE I76 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 1, JANUARYIFEBRUARY 1995 INCOMING L I N E MAIN BUS VOLTAGE :CAE:AKER L i I I I I OPEN FOR HARMONIC C C n \ mn @ o + r: 4 HARMONIC CURRENT ME4SUREMENT P O I N T HARMONIC V O L T A G E MEASUREMENT P O I N T 1 - Incoming Line Voltage a n d C u r r e n t Incoming Main Breaker Voltage and Current 3 - Feeder Breaker Current 4 Low Voltage Substation Main Breaker Voltage a n d Current 5 Harmonic Source Current 6 Capacitor Bank Current ~ ~ ~ Fig. 1. One-line diagram of typical industrial power system showing locations for harmonic measurements. 111. MEASUREMENT POINTS A. Harmonic Filter Design and Capacitor Bank Application The harmonic measurement data needed for design calculations consist of harmonic source currents and voltages. Load equipment, such as rectifiers and variable speed drives, are modeled as harmonic current sources in the calculations. The incoming harmonics from elsewhere in the power system, e.g., the utility interfaces, are modeled as voltage sources. Harmonic current measurements at feeders and at substation main breakers may be used to trace harmonic currents to their sources. Current measurements in a substation should be made at feeder breakers in preference to the main breaker (see Fig. 1 and 2). If some of the feeders go to impedance loads, such as motors, then measurements should be made on the feeder(s) of interest. The reason for this is that the harmonic currents will divide between the transformer and the motors. Measurements may be made at the main breaker as a check. The sum of the feeder harmonic currents as measured should not be expected to be the same as the harmonic currents measured at the main breaker, because the phase angles of the currents are not measured. If power factor correction capacitors are present anywhere in the system, they should be disconnected during the measurements to avoid resonance effects. Current measurements should be made on all three phases if possible, because harmonic currents may be unbalanced. Voltage measurements at the point of common coupling are necessary for the harmonic calculations. As with the current data, all three phases should be measured. Fig. 2. One-line diagram of a substation with power factor correction capacitors and dc drives. B. Harmonic Filter and Capacitor Bank Installation Measurements taken after installation do not need to be as extensive as those made earlier. The harmonic content of the capacitor current, bus voltage, and of the current and voltage at the utility tie point should be checked. Harmonic voltage measurements at substation buses will help to locate problem resonances. These measurements should be made with all existing capacitors connected. C. Compliance with Utility and IEEE 519 Requirements Industrial facilities should not supply more than the allowed amount of harmonic current to the utility. The point of common coupling is not always where the metering is located, as is shown in Fig. 3. Due to the expense of high voltage metering equipment, metering is sometimes performed at the secondary level (Fig. 3(b)), and the utility billing is corrected for the losses in the transformer. When harmonic measurements must be made at the secondary level, they too must be corrected. IV. HARMONIC MEASUREMENT TECHNIQUES A. Transducers Current Transformers: It is preferable for safety reasons to make current measurements at locations where current transformers (CT’s) are already present for relaying or metering purposes. A clamp-on current probe is normally attached to the CT secondary leads. This may be done at the back of a relay, or at ammeter or ammeter switch. CT accuracy is usually specified only for 60 Hz. According to the IEEE Standard 519-1992 [l], CT’s have an accuracy in the range of 3% at frequencies up to 10 kHz [ 2 ] .The frequency response characteristics of a CT circuit are a function of the internal impedances of the CT and its burden. This is illustrated in Fig. 4 and 5. More work needs to be done on the accuracy of CT circuits at harmonic frequencies. The impedances are defined as follows: Primary Impedance. This is not a factor in CT 2, error. SUTHERLAND: HARMONIC MEASUREMENTS IN INDUSTRIAL POWER SYSTEMS INCOMING LINE P U I N T OF COMMON CDUPLING ~ 9 L A I I T Le ICs Lb (a) (b) INCOMING LINE Fig. 4. Current measurement (a) circuit and (b) equivalent circuit [2]. I POINT O F COMMON COUPLING CT RATIO CORRECTION FACTOR 1.01 - /* 1.m.~ 1.008 ~. 1.007 .. With Burden 1.03 -- 8 1.005 -- ~ U 1.004 -1 . m -- c 4- WilhoutBurden / ~ - - - - / / / 1.m ID31 -: -- 1. i ~ ~ ~ ~: ~, , ~ Fig. 3 . One-line diagrams showing point of common coupling: (a) primary metering and (h) secondary metering. ZCS 2, Secondary winding capacitance. Excitation Impedance. This frequency dependent inductance forms a parallel resonant circuit with zcs . Secondary winding impedance. This is smaller than Z,, resulting in a very high resonant frequency. Burden Impedance. This may have frequency zb dependence as well, which is not modeled here. The Ratio Correction Factor may be calculated from the formula [2]: 2, RCF = 1 + 2 s + zb + -2..g + zb ~ zcs z e (1) The resonant frequency of 2, and ZCSis well above the range of interest. If resistance is neglected, the resonant frequency of L and C is: 1 f=T 2" If the inductance is frequency dependent, L L = O d7 (3) then the resonant frequency becomes: (4) For values of LO = 8, and C = 1 nF for a CT tested by Douglass [2], the resonant frequency is approximately 21.6 kHz. This is well above the range of interest. The percentage error as compared to the fundamental may be calculated using (1). For the CT [2] and burden combination shown, the maximum error at the 50th harmonic of 60 Hz is less than 1.0%. This is illustrated in Fig. 5 . If similar data was available for other CT's, and if burden impedances versus frequency were known, the accuracy of actual measurement situations could be evaluated. 178 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 1, JANUARYEBRUARY 1995 for easy attachment to the circuit being investigated. A small burden resistance (typically 5 R) is attached to the CT output. It is selected so as not to cause saturation, but to provide sufficient input voltage for the instrument. Frequency response specifications and/or curves should be available I J from the manufacturer to aid in selection. Clamp-on CT’s are available in many styles, current ranges, frequency ranges, and accuracies. Many CT’s designed for use at 60 Hz have an accuracy that falls off rapidly at the higher harmonics, while CT’s designed for higher frequencies may not be accurate at 60 Hz. Suitable current probes have an accuracy of 0.5 to 2% over a frequency range of 50-10,000 Hz. Hall Effect Probes: Hall effect current probes are similar to clamp-on current transformers in their manner of usage. The semiconducting Hall effect device allows dc currents to be measured as well as ac currents. The specified accuracy for Icput Le Zcs z m Zb CJrcult typical commercial probes is 2 to 5% over a frequency range (b) of dc to 500 or 1000 Hz. Hall effect probes have not been Fig. 6. Voltage transformer, burden and harmonic analyzer input: (a) circuit generally used for harmonic measurements in power systems. and (b) equivalent circuit [3]. Voltage Probes: Voltage probes used for harmonic measurement serve two functions: to reduce the measured voltage Voltage Transformers: Magnetic type voltage transformers to one unsable by the instrument, and to minimize disturbance (VT’s) are typically used in Industrial Power Systems. The to the system being measured. Standard oscilloscope probes typical accuracy of a VT [l] is within 3% for harmonics less generally have been used for harmonic measurements. The than 5 kHz [3]. A typical VT measurement circuit is shown two major characteristics of an oscilloscope probe are high resistance (typically 10 MR) and low capacitance (typically in Fig. 6. The Ratio Correction Factor (RCF) [3] is: 10-100 pF for a l o x probe and 2 pF for a lOOx probe) [4]. The inductance is negligible. Probe frequency response is in the hundreds of MHz. Care should be taken not to exceed the where voltage rating of the probe. Winding Impedance. This includes a series R-L The ground lead of the probe is a critical part of the meaZ,, term, and a primary-secondary capacitance. surement circuit. If a differential input amplifier is available, z b Burden Impedance. two probes may be used, and the ground leads removed from ZCS Secondary capacitance to ground. the probes. Where a single-ended measurement must be made, Excitation Impedance. This varies with frequency the ground lead may be connected to the grounded side of the Ze in a similar manner to the CT impedance. circuit being measured only i f the voltage is first measured Impedance of the measurement circuit. Z, between the instrument’s ground (usually the case, which is The input circuit consists of the resistance and inductance of connected to the third wire of the 120 V plug for safety) and the system forming a low-pass filter with the input capacitance the point where the ground lead is to be attached, and found of the VT. This does not affect the response in the frequency to be within acceptable limits. It is often the case in a power range considered here. The internal resonant frequencies of system that the ground of Voltage Transformers in switchgear the VT may be analyzed in a similar manner to those of the is not the same as the ground of nearby receptacle outlets. If CT. These resonance also are above the frequency range of these two grounds are not the same, applying the ground lead interest for harmonic measurements. may cause a short circuit to occur, which may be dangerous Direct Connections: Direct connection may be made to to personnel and damaging to equipment. Low Voltage circuits in cases where CT’s and VT’s are Newer instruments use special purpose voltage probes denot available. Voltage measurements may be performed using signed for power system use. a 100:1 Oscilloscope probe with sufficient voltage rating. Current measurements may be made by attaching a clampon Current Probe around a cable or busbar. For all direct C. Instrument Input Integace measurements, the connection should only be made while Impedance: The input impedance of an instrument is the circuit is de-energized, locked Out and grounded. The characterized by a resistance and a capacitance. Typically, grounds must be removed, and the circuit re-energized before harmonic analyzers have a l - M ~input resistance that is the measurement is made. compatible with scope probe inputs. The input capacitance of B. Probes Clamp-on Current Transformers: The clamp-on transformer contains a split-core in a hinged structure that allows an instrument may be in the range of 10-100 pF. This is too small to have a noticeable effect for harmonic measurements. If a probe is used, its input resistance and capacitance apply, not those of the instrument. 179 SUTHERLAND HARMONIC MEASUREMENTS IN INDUSTRIAL POWER SYSTEMS Voltage: The inputs of instruments are only able to withstand a small voltage. Overvoltages will cause measurement errors and possibly equipment damage. Because of this, the use of a lOOx oscilloscope probe is recommended. The input range selection should be carefully noted. Whether the input has manual or auto-ranging, the best accuracy is obtained when the amplitude signal being measured is near the top of the range being used. D. Harmonic Analyzers Types: Harmonic measurements may be made with any of several types of measurement systems. Almost all harmonic analyzers now use the Fast Fourier Transform (FFT) calculation procedure 191 for sampled data. Spectrum Analyzers combine the functions of data input, calculation, and presentation of harmonic data, usually on an oscilloscope screen. These are available in both laboratory and portable types. A wide variety of frequency ranges and accuracy levels are available, some of which are suitable for power system use. Portable Computers may be fitted with analog input boards and used as a spectrum analyzer. The FFT may be performed in software or on a digital signal processing board. Power System Harmonics Analyzer. Special purpose instruments are available to measure power system harmonics. Harmonic analysis capability is also available as a feature on other types of instrument, such as power and disturbance analyzers. Digital Storage Oscilloscopes [ 5 ] may be used to gather sampled data, which is then transferred to a computer for processing. Minimum Specijcations: Frequency Range. The frequency range to be measured depends upon the purpose of the measurement. For capacitor and harmonic filter design and application, only those harmonics which have sufficient magnitude to affect the current and voltage rating of the device need to be considered. Typically, measurements may be made up to the 25th harmonic. The requirements of IEEE Standard 519-1992 [ l ] include limitations on harmonics above the 35th order. Instruments measuring up to the 50th harmonic (3000 Hz in a 60-Hz system) may be used. In many situations, harmonic magnitudes above the 25th are so low that they have no practical effect. Accuracy. It is recommended in IEEE Standard 519-1992 [ 11, that the harmonic analyzer used have an accuracy of 5% of the harmonic limit specified in that standard. For example, the limit for current harmonics supplied from a distribution system at the point of common coupling is 0.3% of the fundamental for odd harmonics of order greater than 35. An error of 5% in measuring 0.3% means that the overall accuracy must be 0.015% of the fundamental. For even harmonics, the limit is 25% of this, or 0.00375%. In many practical cases, measurement may be limited to the first 25 odd harmonics, resulting in an accuracy requirement of 0.075% of the fundamental. When the instrument accuracy is specified in percentage of a fullscale reading, and the signal being measured is at the upper limit of the measurement range, then these figures may be compared with the instrument’s specifications. Otherwise, the instrument must be more accurate. If an accuracy of 0.015% is required of a signal where the fundamental is 80% of the input range, then the instrument accuracy should be at least 0.012%. Noise in the instrument input circuit may be specified as an additional source of error in harmonic analyzers. This should be well below the level of the input signal. In order to minimize errors caused by noise, the input voltage to the instrument should be well above the lower limit of the input range. If time averaging is available, this can also be used to cancel out random noise. Dynamic Range. [ 101 The dynamic range of the instrument should be checked to ensure that the smallest detectable variations are not larger than the resolution required. The binary word length of the Analog to Digital Converter (ADC) provides the dynamic range. In order to resolve a signal variation of 0.015% of full scale, a dynamic range of 76 dB is needed. This may be provided by a 14 bit ADC. ADC specifications also include a linearity error on the order of f l Least Significant Bit. This ADC error is included in the overall instrument specification. Aliasing occurs when signals of frequencies higher than the upper limit of the spectrum displayed are sampled and included as if they were lower frequency spectral lines. Aliasing may cause measurement errors when signals are present in the system which have a higher frequency than the highest frequency being measured. This effect is minimized by anti-alising filters, present in all harmonic analyzers. Because aliasing is caused by outside signals, its effect cannot be included in the percentage error of the instrument. The specification which describes aliasing is the steepness of the anti-aliasing filter. Aliasing is a consequence of the sampling process, and operates on the same principles as the sampling of a frequency within the range desired. For an N point sample, frequencies from 0 to N / 2 - 1 cycles per sample period are captured, and the FFT will display them at the correct frequency. When the signal frequency is between N / 2 and N cycles per sample period, two or fewer samples per cycle are taken Aliasing occurs when the FFT maps these frequencies into the range of 0 to N / 2 - 1. The amplitude level of the signals, so that a signal of frequency 1 is mirrored at N - 1. When the FFT is calculated, the magnitudes of the lines from 0 to N / 2 - 1 are doubled, and the higher order lines are dropped. Let the sampled signal be expressed as a function of the sample number, n: f ( n ) = sin where h is the harmonic order, N is the number of samples, and 0 is the phase angle of the signal. If h is expressed as a function of N / 2 and a, where a is the order of the aliasing: IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 1, JANUARY/FEBRUARY 1995 180 Aliasing o f 6 5 t h H a r m o n i c VT RATIO CORRECTION FACTOR 1 5 9 13 17 21 25 29 33 37 41 45 49 128 S a m p l e s HAR MONK (a) Fig. 7. Voltage transformer, burden and harmonic analyzer input: frequency response curves [3], using the data of Fig. 6. The burden resistance is 2960 R, and the inductance is 1.52 H. Aliasing o f 65th Harmonic 1, This may be expressed as the product of a cosine wave of frequency N/2 and a sine wave of frequency a: f ( n ) = cos ( 7 . I). sin ( 2.1r.n N 2.1r.n.a p) 0.8 .. 1 ._ 0.6 -~ c 0.4 .. (d N 0.2 ~. This product models the modulation effect shown in Fig. 8(a), and may be expressed as two sidebands on N/2: 0 16 32 64 48 80 96 112 Harmonic (b) - s i n 2( .T7 r( . T n -N a) -6)). (9) Fig. 8. (a) Time plot of the effect of sampling a signal of a frequency N/2+1 cycles per sampling period. Here, N = 128 and a = 1. (b) The result of the FFT of this signal, showing aliasing to N / 2 - 1 cycles. The FFT produces both of the above terms in an N-line output. An example is shown in Fig. 8(b). This is reduced to N/2 lines, with the values doubled, in the final output. If a is negative, then the signal is within the range of the FFT and the aliasing is at the higher frequency of N - a cycles. A harmonic analyzer that samples 128 times, as in Fig. 8, may have a display that goes to the 50th harmonic. The 78th harmonic would be aliased to the 50th with a magnitude 1/n of 0.0128, causing a 64% error in the measurement of the 50th harmonic of magnitude 0.02. Anti-aliasing filters are used to reduce this error. 50 60 70 80 90 100 110 120 Anti-aliasing filters are steep rolloff analog or digital filters H a rmonic placed in the instrument before sampling takes place. The filter rolloff starts below the frequency N/2. Typical starting points Fig. 9. Anti-aliasing filter rolloff at -130 dB/decade starting at the 50th are 50 out of 64 and 400 out of 512 cycles per sampling harmonic. For a @-point FFT,harmonics beyond the 78th would be aliased. period. A typical value of rolloff is -130 dB per decade. For the example of Fig. 8, the attenuation would be: frequency that can appear in the display is 3120 Hz, or the 52nd harmonic. At -130 dB/decade starting at 2000 Hz, the Atten. = 1301og attenuation at 3120 Hz is 25.1 dB, as in the example above. As above, this will result in a 3.5% error in the 33rd harmonic = 25.1 dB for l / h amplitude distributions. = 5.56%. Windowing is used in spectrum analyzers to provide filtering Thus the aliased signal is .0556/78 = 0.000712 per unit, in the time domain and bandwidth control in the frequency causing a 3.6% error in the 0.02 per unit 50th harmonic. The domain. Windowing in an FFT spectrum analyzer performs slope of this anti-aliasing filter is illustrated in Fig. 9. the same function as bandpass filtering did in earlier spectrum For another example, suppose that the spectral lines are analyzers which used analog electronics. In the time domain, spaced every 5 Hz, and 400 lines are displayed, the maximum discrepancies due to changes in the signal during the sampling harmonic shown is the 33rd, and the maximum frequency period are removed. Windowing is performed by multiplying is 200 Hz. With 512 b in the FFT, samples are taken up the sampled data by a window function, which is zero at each to 2560 Hz, just below the 43rd harmonic. The first aliased ' end of the time record. ):( SUTHERLAND: HARMONIC MEASUREMENTS IN INDUSTRIAL POWER SYSTEM[S A signal should have exactly the same voltage at the beginning of each consecutive sampling interval [8]. This may be accomplished if the system is in a steady-state condition with only integral order (characteristic) harmonics and the sampling interval, T , is an integral multiple of the fundamental cycle time, l / f . Under these conditions, each spectral line of the FFT contains information on the magnitude of its respective harmonic only. For N samples, N / 2 lines are generated, giving a maximum harmonic order of 181 Third Harmonic x 1.01 1 0.8 0.6 0.4 0 :. 0.2 0 3 -0.2 -0.4 -0.6 -0.8 1 If 1024 samples are taken in 0.2 s, h,, = 42. The spectral lines are spaced U0.2 or 5 Hz apart. In order to detect noncharacteristic harmonics, a frequency resolution of less than 60 Hz is required. The amount of resolution between characteristic harmonics depends upon the length, T , of the sampling interval. The bandwidth of the measurement is usually wider than the resolution due to windowing. The N samples of the waveform being measured are assumed by the FFT to be exactly repeated over all time. If there is a discrepancy in the endpoints, then errors will occur where signals from the one frequency are added to those at another frequency in the result. This is called spectral leakage. Suppose that the sampled signal has a small frequency error, E , in comparison with the sample period, T : f ( t )= sin(w. t . m(1 + E ) ) 03 C, cos(w . t . n + 4,) where = Abs 2ns::m~) + 0 1 2 3 4 5 6 7 8 9 10111213 Harmonic Order (b) Fig. IO. Illustration of spectral leakage. (a) Third harmonic signal with 1% frequency error. (b) Magnitude of leakage for lower order harmonic frequencies. (13) n=O { magnitude Third Harmonic x 1.01 showing leakage. (12) where w = 2 . K . f , and m is an integer. The Fourier Series of this function is: ~ ( t=) 128 Sampler + + and a = m ( l E ) - n, b = m ( l E ) n. Notice that m is the original harmonic order of the measured signal, while n is the harmonic order of the calculated harmonic. This equation may be used to calculate the leakage error for a small frequency deviation between the frequency being measured and the sampling rate of the instrument. As an example, a third harmonic signal measured with d frequency error of 1% gives a second harmonic component as follows: m = 3, n = 2, E = 0.01, a = 1.03,b = 5.03, Cn = 0.023. The second harmonic leakage component is thus 2.3% of the third harmonic signal. This example is illustrated in Fig. 10. If the error, E , goes to zero, then C, also goes to zero unless m = n. When m = n, sampling period to the power system frequency. This is valid as long as neither the frequency nor the harmonic content change during the sampling interval. In the general case, the approach that is used is to multiply the data by a window function that forces it to zero at each end of the sampling interval [6], [7]. This can only be done accurately if several cycles of the fundamental are included in the sample interval. Twelve or more cycles are usually taken. The window function that provides the best measurement accuracy is the Flat Top or P201 window, developed by R. W. Potter [6]. This window has an accuracy of 1% or better for signals within the passband, and has minimal ripple outside, resulting in very little leakage. The lack of a window is a window itself, called the uniform window [8]. The passband of the uniform window is the same shape as the Fourier Transform of a rectangular pulse. This is called the Dirichlet kernel, and crosses the horizontal axis at integral multiples of 1/T. All of the signals within any of the lobes are reported in the FFT as part of the harmonic being measured. E. Output Formats which goes to 1 if E goes to zero. One method of making m = n is to use a phase-locked loop circuit to lock the Data Storage Requirements: Because of the large amount of measurement data collected, the harmonic analyzer should have sufficient nonvolatile data memory. Bubble memory was IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 31, NO. 1, JANUARYEBRUARY 1995 182 used in earlier analyzers. Some newer instruments use removable solid-state memories called PCMCIA (Personal Computer Memory Card International Association) cards. Many of these use “Flash” memories, a form of electrically erasable programmable read-only memory (EEPROM). Memory cards which use RAM, on the other hand, contain a small battery. Floppy disk data storage may be used if the environment permits. Printed Results: The measurement results may be printed with both time and frequency plots alongside the tabulated harmonic magnitudes, as shown in Fig. 11. These results may be printed automatically by the use of a computer or printer which interfaces with the harmonic analyzer. A STORE0 RANGE RAEKON-A B STORED TREKON-A -51 OBV STATUS I ”Jolt 5 m V o 1t /DIV I -20 v . EVALUATION OF RESULTS PAUSED X 6 0 HZ Y 9 644 m v r m s A. Additional Calculations Needed The result of a harmonic measurement consists of a millivolt level for each frequency on each phase. This can be converted to system volts or amperes by multiplying by the CT or VT ratios and by taking into account transducers and burden resistors. It is important that all necessary data be recorded in the field. A typical data recording form is shown in Fig. 12. This also contains spaces for recording data directly if an instrument with memory is not available. If the power system is to be modeled with a one-line equivalent circuit, the harmonic magnitudes from all three phase should be averaged. The harmonics should be converted to a percentage of the fundamental, as well as to system amperes and volts. The total harmonic distortion (THD) and total demand distortion (TDD) [l] must also be calculated. Harmonics are evaluated on a steady-state basis, meaning that design values are constant for periods of one hour or more [l]. Thus it may be necessary to average several measurements taken at different times. This may be done within the instrument, or later when the data is being analyzed. Careful measurement procedure should eliminate this added complication in most cases. 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00 17.00 18.00 19.00 20.00 21 .oo 22.00 23.00 24.00 25.00 26.00 27.00 28.00 29.00 30.00 31.00 32.00 33.00 RSS B. Use of Results THD Vsurn - = = 60 120 180 240 300 360 420 480 540 600 660 720 780 840 900 960 1020 1080 1140 1200 1260 1320 1380 1440 1500 1560 1620 1680 1740 1800 1860 1920 1980 9.643840 0.172385 0.111622 0.031441 2.228435 0.010796 0.406912 0.031028 0.160509 0.021 196 0.838977 0.014817 0.247742 0.050962 0.165291 0.020547 0.516714 0.014085 0.126692 0.021903 0.149194 0.016943 0.320812 0.019600 0.102153 0.039994 0.151095 0.023227 0.274789 0.017199 0.022349 0.021852 0.119536 100 .oooo 1.7875 1.1574 0.3260 23.1073 0.1119 4.2194 0.3217 1.6644 0.2198 8.6996 0.1536 2.5689 0.5284 1.7140 0.2131 5 :3580 0.1460 1.3137 0.2271 1.5470 0.1757 3.3266 0.2032 1 .0593 0.4147 1.5668 0.2409 2.0494 0.1783 0.2317 0.2266 1.2395 9.9768 mV 26.5054 X 16.1146 mV The harmonic analysis program should first be used to (b) model the existing system as a base case. As many case Fig. 1 1 . (a) Example of harmonic time and frequency plots. (b) Example of conditions as necessary should be used to model all switching harmonic measurement results in tabular format. configurations in the system. From this the maximum harmonic current magnitudes can be determined for the PCC and compared with the limits specified in IEEE Standard 5 19-1992 should be done for all case conditions, and for all permutations [ 11, or other standard as required. This will aid in determining of component tolerances within each case condition. When appropriate capacitor and/or filter sizes and locations whether a harmonic filter is necessary. The harmonic currents and voltages at existing capacitors and harmonic filters may have been determined, the current and voltage calculations performed originally should be repeated. Using the design values, be calculated and compared with their ratings. The next step is to analyze the system for resonances, using currents and voltages should be calculated for all permutations existing or recommended capacitors as necessary. Calculations of component tolerance during all case conditions. The worst performed usually include determining amplification factors case results should then be used in specifying the components. for injected currents, and impedance magnitudes and angles. The current values at the PCC should be compared with the Both of these calculations produce plots of harmonic scans limits given in IEEE Standard 519-1992 [l], or other required across the frequency range of interest. These calculations standard, and with base case values. The harmonic currents 183 SUTHERLAND HARMONIC MEASUREMENTS IN INDUSTFUAL POWER SYSTEMS When the measurements have been completed, they may be used in harmonic calculations to design filters and to check compliance with standards. If the measurements have been performed with their final use in mind, the final result will be both accurate and useful. ACKNOWLEDGMENT The author would like to thank Randall Schlake of General Electric Company, who developed some of the test procedures described here, including the document shown in Fig. 12. The author would also like to thank Louie Powell of General Electric Company for his many helpful suggestions. REFERENCES , I I I Fig. 12. Example of a harmonic measurement data collection form. should be reduced to an acceptable level in the final result. VI. CONCLUSION By careful selection of equipment and techniques, useful harmonic measurements may be made in industrial power systems. The accuracy of the various portions of the measurement system may be estimated as follows, in percentage of the measured harmonic: Instrument Transformers 0.1-1% Current Probes 0.5-2% 5% Harmonic Analyzer Aliasing 0 - 3.6% Windowing 1-1 % Leakage 0-5%. Overall measurement error should be on the order of 10% of the harmonic magnitude. If a harmonic is measured which is 0.3% of the fundamental, the overall accuracy required is 0.03%. The instrument accuracy should then be 0.015% of the fundamental or better. The least expensive way to improve accuracy is to use the spectrum analyzer function selections wisely. Current probe selection is also of critical importance. The use of probes not designed for use above 60 Hz can result in far greater inaccuracies. View publication stats IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems, IEEE Standard 519-1992, 1993. D. A. Douglas, “Current transformer accuracy with asymmetric and high frequency fault currents,” IEEE Trans. Power Apparatus and Syst., vol. 100, no. 3, pp. 1006-1011, March 1981. -, “Voltage transformer accuracy at 60 Hz voltages above and below rating and at frequencies above 60 Hz,” IEEE Trans. Power Apparatus and Syst., vol. 100, no. 3, pp. 1370-1375, March 1981. ABC’s of Probes, Tektronix, Inc., Beaverton, OR, 1989. J. K. Winn, Jr. and D. R.Crow, “Harmonic measurements using a digital storage oscilloscope,” IEEE Trans. Ind. Applicat., vol. 25, no. 4, pp. 783-788, July/Aug. 1989. Dick Benson, Techniquesfor Signal and System Analysis, PN 363-010001, Tektronix, Inc., Beaverton, OR, 1991. The Fundamentals of Signal Analysis, Application Note 243, HewlettPackard, Palo Alto, CA, 1985. F. J. Harris, “On the use of windows for harmonic analysis with the discrete Fourier transform,” in Proc. IEEE, vol. 66, no. 1, pp. 51-83. Jan. 1978. W. T. Cochran and J. W. Cooley, “What is the fast Fourier transform?“ in Proc. IEEE, vol. 55, no. 10, pp. 1664-1674, Oct. 1967. “Fundamentals of sampled data systems,” in Application Note AN-282, Applications Reference Manual, Analog Devices, Inc., Norwood, MA, 1993. Peter E. Sutherland (M’83) received the A S . degree in electrical engineering technology in 1979 and the B.S. degree in electrical engineering in 1983 from the University of Maine, Orono. In 1986, he received the M.Sc.E. degree in electrical engineering from the University of New Brunswick, Canada. He has worked as a Test Engineer and a Disign Engineer for Accutest Corporation, Chelmsford, MA, a manufacturer of automatic test equipment for the semiconductor industry. For a short time, he worked as a Planning Engineer for an electric utility company. In 1987, he joined General Electric Company, and has been employed as an Engineer in the Industrial Power Systems Engineering Operation, Schenectady, NY, as an instructor in the Training and Development Center, and as a Power Systems Engineer in Albany, NY. He is currently enrolled as a part-time student In the Ph.D. program in electric power engineenng at Rensselaer Polytechnic Institute, Troy, NY. Mr. Sutherland is a member of Eta Kappa Nu and Tau Beta Pi. He is a Registered Professional Engineer in Maine and New York.