Sunteți pe pagina 1din 6

Implementation of Low Cost Non-DFT based Phasor Measurement

Unit for 50 Hz Power System


Talha Ahmed Bhatti, Abdur Raheem, Tanveer Alam, Muhammad Owais Malik, Abdullah Munir
Department of Electrical Engineering
NED University of Engineering and Technology
Karachi, Islamic Republic of Pakistan
talha.bhatti@yahoo.com

Abstract- The interconnection of several small & large systems extracted from the data in real time using statistical analysis
has increased the probability of large blackouts in the power tools. Furthermore control commands are generated depending
grid. One of the leading reason behind these blackouts is low whether the estimated parameters are within prescribed range
reporting rate of existing devices and unavailability of
synchronized system wide data. Synchrophasor technology
or not. The data & information flow of synchrophasor network
provides solution to the problem. Phasor Measurement Unit is depicted in Fig. I.
(PMU) is the most front end device required to implement
Synchrophasor Technology. The paper proposes a non-DFT
~
GPSSatellite
based technique for development of PMU conformed to IEEE
standards C37.118.1 and C37.118.2. The proposed technique
aims to reduce the processing time & sampling rate which
~
ultimately leads to reduced cost. The hardware of PMU is ~
implemented using Raspberry pi Model B+, ADS 1015 and GPS
(Global Positioning System) model PA6H on MTK3339 chipset Data analyze

Make decision
Control system

Keywords - IEEE C37.118, Phasor Measurement Unit,


PMU, Synchrophasor

I. INTRODUCTION

T HE power system of Pakistan is facing severe blackouts


which causes great economic and life losses. The main Fig. I Synchrophasor infonnation flow architecture
reason behind these blackouts is the unavailability of system
wide synchronized data and high latency of the monitoring
system. When there is a fault in one part of system or if any II. PHASOR M EASUREMENT U NIT IMPLEMENTATION
major generating unit is cut off from the national grid, then the
remaining generators also begin to trip one by one due to The heart of phasor measurement unit is the measurement of
excessive load stresses. This catastrophic failure results in a time stamped phasors synchronized to a common time source
system wide blackout. The blackouts can be eliminated by called Synchrophasors. [I] Synchrophasor can be estimated
making the monitoring system quick and wide using using different algorithms on the sampled data. The choice of
synchrophasor smart grid technology. For example if any line algorithm depends primarily upon three factors 1) Accuracy 2)
becomes faulty or generation reduces due to any fault, the Processing complexity (clock cycles required) 3) Response
smart control station isolates the faulty section and reduces the time. [2]
load in proportion with the generation within no time and The existing algorithms include the DFT (Discrete Fourier
thereby avoiding the system wide load stresses. Transform) & its improved version FFT (Fast Fourier
Synchrophasor Smart Grid technology can contribute to Transform) [3], weighted least square estimation [4] and
overcome several currently prevailing problems of power discrete wavelet transform methods [5] . The methods require
system e.g. blackouts, low reporting rate, unavailability of high sampling rates to avoid aliasing. It has been observed that
universally synchronized and time stamped data. The many of these references are prone to the spectral leakage &
technology aims to improve existing supervisory control and interference phenomenon. Although several techniques are
data acquisition system (SCADA). The prime differentiator used to mitigate the effect but it further increases the overall
between conventional and smart grid is the use of processing complexity of the algorithm [6].
synchrophasor data. The IEEE standard C37.IIS.1 does not require calculation
The synchrophasor architecture consists of a network of of higher order harmonics of the given signal i.e. only 1st
PMUs whose output data is concentrated using Phasor Data harmonic or fundamental component is required which can be
Concentrator (PDC). There can be several stages of data calculated using DFT based algorithm. Nyquist criteria will
concentration i.e. local, regional then super PDC. The suggest sampling rate of 100 Hz or above for 50 Hz nominal
concentrated data is then transmitted to monitoring & control frequency system, but in order to avoid spectral leakage that
center. Then several important electrical parameters are can occur due to the non-integer frequencies (non-multiple of
50 Hz) present in the signal a high sampling rate will be

978-1-5090-1252-7/16/$3l.00 2016 IEEE


required along with large window size. But large window will time synchronization unit, is implemented on microcontroller
increase the latency i.e. response time & also the PIC (Peripheral Interface Controller) 16f877a. The Linux is
computational requirement of the algorithm. Therefore, not a RTOS (Real Time Operating System), so it cannot be
fabrication of Non-DFT based PMU is represented in this used for precise timing operations. Therefore, the sampling
research paper. operation and transmission of message frames is dictated by
The advantage of using this non-DFT based phasor the ADPLL's output. The ADPLL's output can be a 10, 50 Hz
estimation technique is that the required sampling rate of PMU or any other higher frequency digital signal. The ADPLL's
is very low as compared to DFT & other techniques. Required frequency is set as per the required reporting rate ofPMU. The
sampling rate is equal to the desired reporting rate using this ADPLL's output wave is synchronized with 1 PPS signal, so
technique i.e. for 10 Hz reporting rate, the sampl ing rate is 10 that the sampling and transmission operations are in
sps (samples per second) & for 50 Hz, it is 50 samples per synchronism with the 1 PPS signal.
second & so on. Hence, there is no need to use high speed GPS is a part of time synchronization block. Two type of
ADCs. Considering 10 Hz rate, the processing power of the outputs are taken in the PMU development technique
microprocessor also reduces significantly because now only discussed in this paper, one is 1 PPS signal & other is UTC
10 samples are required to be processed. Hence, a very low (Co-ordinated Universal Time) time. UTC time is sent to the
processing power microprocessor and low cost ADC is Raspberry Pi using UART (universally asynchronous
required. receiver/transmitter). The standard defmition of UTC is: "The
time of day at the Earth's prime meridian (0 longitude)". This
III. FUNCTIONAL BLOCKS OF PMU UTC time is converted into SOC (Second of Century) for time
stamping the phasors conformed to IEEE standard.
The major functional blocks ofPMU are depicted in Fig. 2:
IV. MAGNITUDE ESTIMATION
Sa tellites
Time
The fundamental component of signal is to be calculated
'--_+I Synchronization only. The rms value of the given signal is estimated using ac
10 Hz to dc rectification. The ac signal is passed through a bridge
UTCTime
SIgnal From Cl ienL rectifier to convert it into dc. The dc signal is then a reflection
,-----,
of the given ac waveform. The ac value is obtained by
Signal Data multiplying the output of rectifier by a pre-calculated factor
Data
Acqui sitio n &
Processing Commu nication Therefore, the magnitude of a signal can be measured by
Co nditi oning
simply measuring the dc value using ADC. The magnitude
ToClienL measurement will not be effected by any frequency change,
i.e. the magnitude measurement accuracy remains almost same
Fig. 2 Functional Diagram of PMU at nominal and off nominal frequencies.

The signal conditioning block conditions both the input V. PHASE ANGLE ESTIMATION
voltage and current signals. The supply voltage is reduced to
suitable value using power resistors' VDR (Voltage Divide
The phase angle is estimated by calculating the time interval
Rule). The current sensing is done using ACS 712 IC
between the rising edge of the 1 PPS (Pulse Per Second) signal
(Integrated Circuit). Its output is clamped at 2.5 V so a
from GPS & the time of rising of sine wave at zero crossing
capacitor is connected after IC's output to remove this dc
shown in Fig. 4. But the phasor defmition pertinent to
offset. Then these voltage & current phasors are sent to both
C37.11S.l defmes the phase angle of a cosine function so 90
bridge rectifier and hysteresis comparator circuits for further
is subtracted after phase angle calculation using the sine
signal conditioning. Higher order harmonics are filtered from
function .
these signals using low pass filter as shown in Fig. 3. As it can
be seen in Fig. 3 that the low pass filter also introduces a
phase shift in the waveform but this phase shift is constant.
Therefore, this additional phase shift is subtracted from each
estimated phasor to obtain actual phase angle.
t , t I I ,

~
\\!LJ~ ~\)J
Figure 3 Harmonics removal using Low pass filter

All Digital Phase Lock Loop (ADPLL) which is present in Fig. 4. Phase angle estimation
1 1
The sinusoidal input waveform is fIrst converted into square f =B - B D ld = 1'5 (5)
function using hysteresis comparator, so that it can be fed Where, B= SOC time of zero crossing of current cycle
directly to the microprocessor. The output of comparator is a B _old=SOC time of zero crossing of preceding cycle
square wave having low & high voltage corresponding to the
Tp
dual power supply voltages. The microprocessor Raspberry pi
can only accept positive voltages, hence the negative portion
:.. ~I

of the square wave is clipped using a silicon diode. Then the


positive portion is brought down to lower voltage i.e. 3.3 V I-~~--, _ --- _1- _ 1----,,...-- - , ____ ~-
I
using VDR. This positively clamped square wave can now be
given to microprocessor, where the rising edge of the square
wave reflects the zero crossing of sine wave.
The microprocessor is programmed such that the rising
edge generates an interrupt, the time of occurrence of this zero
crossing (rising edge) is measured in the Interrupt service
routine (ISR) & stored in a global variable. In a similar
fashion, the occurrence time of rising edge of 1 PPS signal is
measured and stored in another variable. The two variables
containing the SOC (Second of Century) time are now
subtracted to get the time interval between the PPS & sine
wave zero crossing. This time is then converted into phase B
angle "8 " using (5). For 50 Hz the waveform there are 18000
in 1 sec, hence the time is multiplied by 18000 to convert time Fig. 5. Frequency Estimation
in seconds into angle as shown in (1)
8 = t * 18000 (1) VII. RATE OF CHANGE OF FREQUENCY (ROC OF)
Where, t= time between rising edge of 1 PPS signal & CALCULATION
power signal waveform
Then the angle in (1) is subtracted by 90 because cosine
Another quantity which is still required to be calculated
function is to be considered instead of sine as per standard.
before fIlling & transmitting the PMU output message frame,
8 = (t * 18000) - 90 (2)
is the ROCOF (Rate of change of frequency). It is more prone
Now for further cycles, 360 is to be subtracted as each
to the amount of noise present in the input as compared to
cycle represents 360. Therefore, remainder operation is
other phasor estimates. The rate of change of angle IS
applied on (2) to get (3)
frequency, this relation is used to obtain rate of change of
8 = ((t * 18000) - 90) mod 360 (3)
frequency from rate of change of angle as shown in (6):
Where, mod= remainder operation
Finally (4) is obtained by subtracting 180 from (3) because
0(t)=fO)(t)dt = 00 + ~O)t + 0.5 0)'t 2 (6)
according to the IEEE standard the angle can vary from -180
to + 180, then the degrees are converted into radians.
Where Ts=Time of each sample
~O)= Change of frequency (COF)
8 = (((B- A+0.005)x 18000) mod 360) - 180)x(n/180) (4)
O)'=ROCOF
Where, B=SOC time of zero crossing of sine wave
Writing (3) in matrix form to get (4)
A=SOC time of 1 PPS signal ' s rising edge
In (4), B-A represents the time duration between the IPPS 00 1 0
and the signal' s zero crossing. 01 1 Ts
VI. FREQUENCY ESTIMATION

The frequency estimation can be done on the basis of time


O2
03
1
1
2Ts
3Ts [:~l
X cu'
2
obtained through phase estimation technique as shown in Fig. 0N - 1
5. The two consecutive rising edges ' occurrence time are (7)
subtracted to get the time period. Actually the time interval
between the two rising edges is the time between zero Where 00=Estimated phase angle of fIrst sample
crossings of the sine wave, so just taking the inverse of this 01 =Estimated phase angle of second sample
time gives the frequency of the signal. The fundamental 0 N_l=Estimated phase angle of Nth sample
frequency is calculated using (5) Let N =50 then Ts=0.02 sec
Replacing the matrices in (7) with respective variables
yields (8) frames transmission after it is switched on. The data frame
[0]=[B] [A] (8) transmission is controlled by the command frame, i.e. the data
Where, frame transmission can be halt or resumed by sending
[0] Represents matrix of phase angle estimates in one appropriate command messages. On the other hand the header
complete window & configuration frames are not sent continuously, instead the
[B) Represents coefficient matrix of (8) PMU sends these frames only on the request of the client. One
[A] Represents unknown column matrix of (8) thing to be kept in mind that the PMU is the server & the
PDC, SVP (Synchro Vector Processor) or any other such
The Weighted Least-Squares (WLS) technique is used to synchrophasor data reception device that must be a client as
calculate the unknown vector [A]. Equation (9) is per the IEEE communication standard C3 7 .118.2.
implemented on PMU, in which the matrix G's value remains The header frame is implemented on python Raspberry pi as
constant, whereas matrix 0 is updated containing the latest shown in Table. 1:
estimated phase angles. (9) is derived from (8), whereas (10)
is used to calculate value of matrix "G". Table. 1.Header Frame
SYNC FRAMESIZE IDCODE SOC FRACSEC DATA DATA CHK
2B 2B 2B 4B 4B 3B 2B 2
(9)
OxAA ll OxOO1 5 OxOOOl xxxx xxxx "Tal" "ha" xxxx

Where, "xxxx" represents that the data of this field in not static
[G) = [BTB]-IBT (10) instead its value is different at every instant of time. SOC time
is calculated from the UTC time given by the GPS module
The matrix [G) is stored in real time for use and it is pre built on MTK3339 chipset using UART communication.
calculated. It consists of 50 rows (for 50 sps system) and 3 FRACSEC field is used to indicate the fraction of second, time
columns. The multiplication of matrix [G) with [0] is quality & also informs about leap second insertion or deletion
performed to obtain matrix [A] from which ROCOF can be nearby. Fraction of second is incremented with every rising
found at any time using (11). [3] edge. Its value varies between 0 and 49 for 50 sps reporting
w' rate PMU. The counter resets after completion of each second.
ROCOF =- (11)
21<. The CHK contains the cyclic redundancy check (CRC) of the
whole frame. It is calculated using mkcrcfun module of
python.
VIII. PMU DATA COMMUNICATION
The data frame is implemented as represented in Table. 2:

The data filling, reception & transmission is done as per Table. 2.Data Frame
IEEE standard C37.118.2-2011. TCP!UDP (Transmission
Control ProtocollUser Datagram Protocol) method is the opted
protocol which is most suitable for real time applications. The
modules of python used are socket, struct and RPIO
(Raspberry Pi Input Output). The struct module is used for
handling binary data for network connections. The data is
packed into bytes using struct.pack command, similarly
struct.unpack is used to convert the received bytes into
separate tuple entities as per given format. The socket module
deals with the UDP sockets, whereas RPIO module is utilized
to handle the TCP interrupt sockets. Although socket module The phasor is represented in polar form , while frequency
can also handle TCP sockets but it can't generate TCP and other value are represented in floating point format. The
interrupts, hence RPIO module is used instead of socket data frame is being sent after every 20 ms for 50 sps reporting
module for TCP sockets implementation. TCP/UDP method is rate. There can be several data streams transmitting from a
used for messages transmission. There are four types of single PMU simultaneously but here only one data stream is
message frames according to the standard out of which the implemented, so mCODE is assigned value " 1" which is
three i.e. command, header & configuration frames are going to be constant for all the frames. The actual PMU data
transmittedlreceived using TCP sockets whereas the single frame output is shown in Table. 3. The configuration frame is
data stream message frames are sent through UDP sockets in implemented in the similar fashion .
TCP/UDP method of communication. PMU transmits all the
above mentioned message frames except of the command Table. 3 PMU Data frame output
frame which is received by the PMU. Command frame is SYNC friiiiiiIi. ID soc FRACSEC STAT VOlTAGE ANGLE FRE ROCOF ANAlOG DIG CHK
43521 80 1 1443104668 100 128 268.25 1.89633 50.17 0.1 220 17 48676
usually sent by Phasor data concentrator (PDC) or such any 43521 80 1 1443104669 10 128 268.55 1.07378 50.17 220 17 43521
43521 80 1443104669 20 128 268.29 1.25039 50.17 0.1 220 17 49202
other synchrophasor device pertinent to C37.118.2. [7] 43521 80 1443104669 30 128 267.38 1.43404 50.18 11.56 220 17 49182

The PMU starts the spontaneous transmission of data 43521 80 1 1443104669 40 128 268.21 1.61463 50.18 0.2 220 17 42103
The command frame is different from above frames in the such disturbance
manner that it is to be received & decoded by the PMU. When Power system model validation
a command frame reaches the PMU's port, it generates a TCP Summing up all benefits, it is evident that the value of
interrupt. In interrupt service routine the CMD field of the benefits is much greater than the setup cost. Hence the
frame is interpreted and the rest of the program runs according payback period is equal to the blackout occurrence period.
to it. There can be several commands defined by the user but
there are four basic commands that are defmed by the standard X. CONCLUSION
itself. These are the switching on & off the data frames
transmission and requesting a header or configuration frame The most commercial PMUs estimate several order harmonics
from the PMU. along with the fundamental component, but there are several
applications which only require the data related to
IX. COST B ENEF IT ANALYSIS fundamental frequency component of power system. Hence,
this technique can be used to fabricate very low cost PMUs
The cost of implemented PMU excluding courier and which will reduce the overall implementation cost of
custom charges is around $107 which is much less as synchrophasor network. The actual hardware implemented and
compared to other commercial PMUs, for example the SEL- tested using this methodology is shown in Fig. 6. This PMU is

..
351A's retail price is $1260. The cost breakup of implemented intended to be tested using Phasor Measurement Unit
Calibration System-6135A/PMUCAL or any other such
PMU is shown Table. 4.
testillg equipment.
Table. 4 Cost breakup of implemented PMU
Equipment Cost
Microprocessor Raspberry Pi Model B+ S2S
2 GPS module Model: PA6H, chipset mtk333 S40
GPS Antenna and uFL to SMA connector S17
4 Current Sensing IC ACS71 2T ELC-20A SIS
ADC ADS 10 15 Adafruit S10

Sum $107

Total Cost inclnding conrier charges & custom PKR20000


duty

The cost estimation of overall synchrophasor network is not


straight forward which varies from system to system. It is
mainly due to the variation in installation charges. The
average cost can be estimated as PKR.50000/PMU and
PKR. I00000/PDC which includes manufacturing and
installation charges as well. Assuming that PMUs are installed
in each of the 64 EHT (Extra High Tension- 132 & 220 KV)
grids ofK-Electric's network and a PDC for 10 grids, the total
cost will be around PKR. 3800000. Fig. 6 Prototype of proposed technique
The current GDP (Gross Domestic Product) of Karachi is
around $100 billion. A 12 to 24 hour blackout will cause ACKNOWLEDGMENT
complete industrial shutdown, hence leading to a large We are very grateful to Dr. Muhammad Ali Memon,
economic loss. Apart from monetary loss, there are life losses Chairperson, Electrical Department, for supervising us. We
as well such as power outage in hospitals will halt the are also highly thankful to Sir Hassan-ul-Haque, Lecturer,
emergency operations, surgeries and other such lifesaving Electrical dept. for guiding and helping us in the design,
activities. Similarly the heat stroke death toll during June-20 15 implementation and troubleshooting of various modules of the
climbed past 1200. The effect of heat stroke was increased due hardware.
to the blackouts as these blackouts deprived customers of the
use of air conditioners, fans and other electric equipment. REFERENCES

The benefits of synchrophasor network implementation are


not limited to blackouts prevention instead it provides [1] IEEE Standard for Synchrophasor Measurements for
following benefits also: Power Systems, ill IEEE Std C37.1I8.I-201I
Wide area monitoring which can prevent grid (Revision ofIEEE Std C37.118-2005)201l. p. 1-6l.
collapse [2] Thorp, A.G.P.aJ.S., Synchronized Phasor
Measurements and Their Applications, ed. A.S. M.A.
Enhancing grid throughput and optimization of
Pai. 2008, New York, NY 10013,USA: Springer-
available resources
Verlag.
Quick forensic analysis after blackout or any other
[3] Phadke, A.G., I.S. Thorp, and M.G. Adamiak, A New
Measurement Technique for Tracking Voltage
Phasors, Local System Frequency, and Rate of
Change of Frequency. Power Engineering Review,
IEEE, 1983. PER-3(5): p. 23-23.
[4] Sachdev, M.S. and M. Nagpal, A recursive least
error squares algorithm for power system relaying
and measurement applications. Power Delivery,
IEEE Transactions on, 1991. 6(3): p. 1008-1015.
[5] Chi-Kong, W., et al. A novel algorithm for phasor
calculation based on wavelet analysis (power system
analysis]. in Power Engineering Society Summer
Meeting, 2001.
[6] Romano, P. and M. Paolone, Enhanced Interpolated-
DFT for Synchrophasor Estimation in FPGAs:
Theory, Implementation, and Validation of a PMU
Prototype. Instrumentation and Measurement, IEEE
Transactions on, 2014.63(12): p. 2824-2836.
[7] IEEE Standard for Synchrophasor Data Transfer for
Power Systems. IEEE Std C37.118.2-20ll (Revision
ofIEEE Std C37.118-2005), 2011: p. 1-53.

S-ar putea să vă placă și