,(((3(6$VLD3DFLILF3RZHUDQG(QHUJ\(QJLQHHULQJ&RQIHUHQFH $33((&

State Estimation for Distribution Systems Using

Micro-Synchrophasors
Xuebing Chen King Jet Tseng Gehan Amaratunga
Student Member, IEEE Senior Member, IEEE Department of Engineering
Nanyang Technological University Nanyang Technological University University of Cambridge, UK
[email protected] [email protected] [email protected]

Abstract—Phasor Measurement Unit (PMU) has contributed single phase state estimator for transmission systems no longer
greatly to power system state estimation because of its ability to applicable in distribution systems.
directly measure voltage and current phase angle. PMUs have Several distribution level state estimation methods have
been used almost exclusively in transmission systems monitor-
ing. Recent years, increasing penetration of renewable energy already been proposed to account for imbalances in system
sources brings new characteristics into distribution system such operation. Switching to full three-phase representation of the
as bi-directional power flows and voltage profile issues, which network is the most straightforward method. Decoupled state
has necessitated continuous monitoring of distribution systems. estimation using sequence network could be more efficient by
Therefore, fast and accurate measurement device is needed for decreasing problem dimension [11]. In this paper, a linear three
application in distribution networks. 𝜇PMU is such a device that
can provide phasor measurements with high precision and low phase state estimator was implemented and tested on IEEE
cost. This paper uses a linear three phase state estimator for standard systems. The paper is arranged as follows: section II
applications in distribution systems. The proposed estimator can provides introduction to PMUs and 𝜇PMUs and distribution
make use of synchrophasor measurements which can be realized system state estimation is introduced in section III. Simulation
by 𝜇PMU. This is tested on IEEE 13 bus feeder which has is described in section IV, as well as the result analysis. Section
unbalanced three phase transmission lines and loads.
V concludes the paper and proposes future work.
I. I NTRODUCTION II. PHASOR MEASUREMENT UNITS AND MICRO
Phasor Measurement Units (PMUs) have been used in SYNCHROPHASOR
power system to enhance performance of state estimators. Synchronized phasor measurement units (PMUs) were de-
PMUs are able to provide synchronized voltage and current veloped in the mid-1980s [11]. It can provide synchronized
phasor measurements with high resolution and sampling rate, voltage and current phasor measurements with GPS, which
thus play an essential role in monitoring and control in make it potential to play an essential role in power system
many power transmission networks [1]. The linearity between monitoring, protection and control. Using synchronization
the system states and the measurements allows direct (non- signals from the GPS satellite system, PMUs have evolved
iterative) solution of the state estimation problem. into mature tools and are being manufactured commercially.
Recently years, the increasing penetration of renewable en- Its update rate (30 times per second) and accuracy of synchro-
ergy sources brings new characteristics into distribution system nization (better than one microsecond) enable it to provide real
such as bi-directional power flows, unstable power supply, time measurements and capture the dynamics of power system
unintentional islanding and voltage profile issues [2]. This calls [12], [13].
for the deployment of more accurate and faster monitoring PMUs have been proven as a promising tool to improve
devices [3]. Large amount of work have been carried out in the the performance of state estimation in transmission networks.
application of PMU to distribution system [3]–[7]. However, However, distribution application is more challenging in both
PMU that has been used in transmission networks is not ideal technical and economic aspects: Due to the short transmission
choice for distribution system application due to economic lines and low power flow, voltage angle between locations on
and technical constraints. Micro-synchrophasors(𝜇PMUs) are a distribution network will be up to two orders of magnitude
high-precision measurement units that can work well in dis- smaller than those on the transmission network (tenths of a
tribution networks [8]. degree, rather than tens of degrees) [8]. Moreover, considering
There are lots of potential applications of 𝜇PMU to distri- lower profit and importance as compared to the transmission
bution networks such as phase identification, reverse power setting, the costs must be far lower to make a business case for
flow detection and state estimation [8], [9]. State estimation is the installation of multiple PMUs on a distribution circuit. The
a possible application of 𝜇PMU. Compare with transmission 𝜇PMU developed in [8] is able to discern angle differences
systems, distribution systems have many different features down to ±0.01∘ with low installed cost. Commercial manu-
including radial topology, high resistance to reactance ratio facture of 𝜇PMU will motivate application of synchrophasors
and three-phase unbalanced system [10]. These features make in distribution system.

‹,(((
 III. STATE ESTIMATION The relationship can also be described in matrix form [16].
⎡ ⎤ ⎡ ⎤
𝑉𝑖𝑎 𝑉𝑖𝑎
State estimation is the process that provides the optimal ⎢ 𝑉𝑖𝑏 ⎥ ⎢ 𝑉𝑖𝑏 ⎥
estimate of the current state of the power system using ⎢ ⎥ [ ]⎢ ⎥
⎢ 𝑉𝑖𝑐 ⎥ 𝑰 0 ⎢ 𝑉𝑖𝑐 ⎥
measurements data and network topology information [2]. ⎢ ⎥= ⎢ ⎥ (6)
⎢𝐼𝑖𝑗𝑎 ⎥ 𝒚𝑖𝑗𝑎𝑏𝑐 + 𝒚𝑖0𝑎𝑏𝑐 −𝒚𝑖𝑗𝑎𝑏𝑐 ⎢ ⎥
The state of a power system can be described by complex ⎢ ⎥ ⎢𝑉𝑗𝑎 ⎥
⎣ 𝐼𝑖𝑗𝑏 ⎦ ⎣ 𝑉𝑗𝑏 ⎦
voltage (magnitudes and phase angles) at every bus. State
𝐼𝑖𝑗𝑐 𝑉𝑗𝑐
estimation plays a key role in ensuring the secure operation
of power system. In traditional state estimation, measurement In recently years, distribution networks are evolving into
data are usually voltage magnitude, power flows and power active networks with increasing amounts of distribution gen-
injections. These measurements have nonlinear relationship erators connect to distribution networks. The need for distri-
with power grid states (complex voltages), resulting non-linear bution system state estimation is particularly acute because of
and time consuming state estimator. Non-linear state estimator the bi-directional power flow and more complicated network
can be solved by iterative method, among which weighted least behavior caused by installation of DG [5]. On the other
squares is the most common method [14]. hand, development of 𝜇PMU make distribution system state
estimation feasible on economic and technical constraints.
𝑧 = ℎ (𝑥) + 𝑒 (1) State estimation is a fairly routine task in transmission sys-
tems, but the technique need to be improved to be applicable
In this estimator, 𝑧 is a measurement vector made up with in distribution systems considering the distinctions between
traditional measurements. ℎ (𝑥) is a measurement function the two networks [17]. Transmission networks transfer large
made up of non-linear relationships between the measurement amounts of power through mesh topologies made up with
vector and the state vector x. And e is an error vector. high reactance-resistance ratios transmission line. Distribution
systems typically transfer limited amounts of power through
a radial topology consisting of low reactance-resistance ratio
feeders and unbalanced loads [18]. More 𝜇PMUs are needed
to make the distribution system observable due to the large
amount of nodes and radial topology.
In transmission system state estimation, state estimator is
usually single phase. However, three phase imbalances in
distribution system make the single phase state estimator
no longer accurate. To achieve accurate state estimation for
distribution system, it is suggested to switching to full-phase
representation of network [11]. In three-phase state estimation,
Fig. 1. 𝜋-Model of a transmission line
three-phase bus voltage are designated as the system states.

𝑍 = 𝐻𝑉 + 𝑒 (7)
Mean Absolute Percentage Error (MAPE) and Mean Ab-
solute Error (MAE) can be used to evaluate the accuracy of Measurement vector 𝑍 contains three phase bus voltage phasor
voltage magnitude and phase angle respectively. and branch current phasors. 𝐻 is the measurement Jacobian
matrix. 𝑉 is the system state vector and 𝑒 is measurement error
(∑ ) vector. If all measurements are from PMU, the equation is lin-
1 𝑁 ∣𝑉𝑖𝑒𝑥 −𝑉𝑖𝑒𝑠𝑡 ∣
𝑀 𝐴𝑃 𝐸 = 𝑁 𝑖=1 𝑉𝑖𝑒𝑥 100% (2) ear, which means that H is constant matrix which only depends
(∑ )
1 𝑁 on system topology and measurement configuration. However,
𝑀 𝐴𝐸 = 𝑁 𝑖=1 ∣𝜃𝑖𝑒𝑥 − 𝜃𝑖𝑒𝑠𝑡 ∣ (3)
this approach may be computationally costly because of the
triple dimension. Given that the system is fully observable with
where the upperscript 𝑒𝑥 and 𝑒𝑠𝑡 refers to exact and estimated 𝜇PMUs, that means the state estimation equation is linear and
respectively, and 𝑁 is the number of buses [15]. can be solved directly (non-iterative). In this case, sequence
With the proliferation of synchronous phasor measurements, networks can be deployed to decouple state estimator into
linear state estimator is developed. This is because PMUs three smaller size problems [11]. Let 𝑇 be the transformation
measurements have linear mathematical relations with system matrix which relates phase and sequence networks
states. In a two-port 𝜋-model transmission line (shown in ⎡ 2 4 ⎤
Figure 1), the relationship of current phasor and system states 1 𝑒𝑗 3 𝜋 𝑒𝑗 3 𝜋
𝑇 = ⎣1 𝑒𝑗 3 𝜋 𝑒𝑗 3 𝜋 ⎦
4 2
can be described as follows [3]: (8)
1 1 1
𝑅
( ) ( )
𝐼𝑖𝑗 = 𝑔𝑖𝑗 𝑉𝑖𝑅 − 𝑉𝑗𝑅 − 𝑏𝑖𝑗 𝑉𝑖𝐼 − 𝑉𝑗𝐼 − 𝑏𝑖𝑖 𝑉𝑖𝐼 (4) and 𝑇𝑧 be the block diagonal square matrix having 𝑇 on its
𝐼
( ) ( )
𝐼𝑖𝑗 = 𝑔𝑖𝑗 𝑉𝑖𝐼 − 𝑉𝑗𝐼 + 𝑏𝑖𝑗 𝑉𝑖𝑅 − 𝑉𝑗𝑅 + 𝑏𝑖𝑖 𝑉𝑖𝑅 (5) diagonal. Multiplying both sides of (7) by 𝑇𝑧 from the left
gives: observable. These five 𝜇PMUs are placed at bus 1, 3, 5, 9,
𝑇𝑧 𝑍 = 𝑇𝑧 𝐻𝑉 + 𝑇𝑧 𝑒 (9) 10. The true values were obtained by carrying out a load flow
using Simulink and saved in MATLAB workspace.
Define transformation matrix 𝑇𝑣 as block diagonal square
Since 𝜇PMU measures not only the voltage phasor at the
matrix having 𝑇 −1 on its diagonal. Therefore three phase bus
bus it was installed, but also the current phasor of all incident
voltage vector 𝑉 can be expressed as:
lines, these five 𝜇PMUs returned voltage phasor at five buses
𝑉 = 𝑉 𝑣 𝑉𝑠 (10) and current phasor of nine transmission lines. As we were
developing three phase state estimator, there were 13 voltage
𝑉𝑠 is the sequence component of three phase voltages. Sub- phasor measurements and 20 current phasor measurements.
stituting (10) in(9) yields: Figure 3 shows the 13 bus feeder with bus numbered. Current
𝑍 𝑠 = 𝐻𝑠 𝑉 𝑠 + 𝑒𝑠 (11)
where, 𝑍𝑠 = 𝑇𝑧 𝑍, 𝐻𝑠 = 𝑇𝑧 𝐻𝑇𝑣 and 𝑒𝑚 = 𝑇𝑧 𝑒. (11) can be
expanded as:
⎡ +⎤ ⎡ ⎤ ⎡ +⎤ ⎡ +⎤
𝑍𝑟 𝐻𝑟 0 0 𝑉𝑟 𝑒𝑟
⎣𝑍𝑟− ⎦ = ⎣ 0 𝐻𝑟 0 ⎦ ⎣𝑉𝑟− ⎦ + ⎣𝑒−
𝑟
⎦ (12)
𝑍0 0 0 𝐻0 𝑉0 𝑒0
From this equation, it can be seen off-diagonal elements are
zeros, so this matrix equation can be written as three separate
equations and solved one by one. Hence the computational
time can be reduced greatly with help of parallel processors
[11].
IV. SIMULATION RESULTS
In this section simulation was conducted to evaluate the
three phase state estimator on distribution system. IEEE 13 bus
feeder was selected to be the test feeder. The distribution sys-
tem was modeled using MATLAB/Simulink. The codes were
written using MATLAB and implemented on a computer(Intel Fig. 3. Test Feeder with Measurements
Core i7)with processor speed of 2.4GHz.
Figure 2shows the topology of the selected test feeder.
and voltage measurements are also indicating in this figure.
The blue boxes denotes phasor measurements, reflecting the
placement of 𝜇PMUs. Current measurements are located at
incident lines and denoted by red boxes.
There were some simplifications in the simulation. Firstly,
with such configuration of 𝜇PMU, most states of the system
were observable except for phase B of bus 7 because transmis-
sion line connected bus 7 and 8 was two phase (phase A and
phase C). Hence phase B of bus 7 was not included in the state
estimation. Note that bus 7 and bus 8 were directly connected
through a three phase breaker. Phase B of bus 8 was assigned
to that of bus 7. Secondly, no load at bus 12 determined no
current flow through transmission line connected bus 7 and
bus 12. Thus three phase voltage at bus 12 must equal to that
of bus 7.
Therefore, number of states that to be determined was
28. With these 33 measurements, 28 states were estimated.
Bias were added to all measurements. According to specifi-
cation [20], guaranteed amplitude accuracy is ±0.05%, and
guaranteed angle accuracy is ±0.01∘ . Bias that added to
Fig. 2. IEEE 13 bus feeder [19]
measurements were random number within the given accuracy.
State estimation equation was developed follow this proce-
In order to develop linear three phase state estimator, the dure.
network must be fully observable by 𝜇PMUs. For this test 1) Voltage Measurement Bus Incidence Matrix 𝑰𝑰
feeder, five 𝜇PMUs were necessary to make all bus voltage Each row of this matrix corresponds to voltage measure-
 ments, and column number depends on system states. Phase B Voltage Magnitude Phase B Voltage Phase
1 −117
Elements of voltage Mmasurement bus incidence matrix True True
is either 1 or 0. If measurements 𝑋 is located at bus 𝑌 , 0.99 SE −117.5 SE

the matrix element (𝑋, 𝑌 ) will be a 1. All remaining 0.98 −118

entries will be 0 [21].
2) Current Measurement-Bus Incidence Matrix 𝑨 0.97 −118.5

Voltage Phase (degree)

Voltage Magnitude (pu)
Row and column numbers os this matrix are determined 0.96 −119
by current measurements and system states. If mea-
0.95 −119.5
surement 𝑋 leaves (injects) bus 𝑌 , the matrix element
(𝑋, 𝑌 ) will be 1 (−1) [21]. 0.94 −120
3) Series Admittance Matrix 𝒚 0.93 −120.5
This matrix is a square matrix. Off-diagonal elements
are zero and diagonal elements are the admittance of 0.92 −121

each measured branch [21]. 0.91 −121.5

4) Shunt Admittance Matrix 𝒚𝒔
0.9 −122
This matrix has exactly dimension with Curren- 0 5 10 0 5 10
t measurement-bus incidence matrix. If measurement 𝑋 Bus number Bus number

leaves bus 𝑌 , the matrix element (𝑋, 𝑌 ) is the shunt

admittance of the side of the line where measurement Fig. 5. Phase B estimation results and true value
𝑋 was taken [21].
[ ] [ ] Phase C Voltage Magnitude Phase C Voltage Phase
𝑽 𝑰𝑰 1 120
[𝒛] = = [𝒙] + [𝒆] (13)
𝑰 𝑴 0.995
True
SE
True
SE
119
0.99
𝑴 = 𝒚𝑨 + 𝒚𝒔 (14)
0.985

Voltage Phase (degree)

Voltage Magnitude (pu)

118
Since MATLAB could directly solve complex equations, the 0.98
state estimation equation was solved without separating real
and image part. It was also validated that using real and image 0.975 117

separately gave exactly the same results. The resulting state 0.97
estimation for phase A is compared with true value in Figure 4. 116
0.965

0.96
115
Phase A Voltage Magnitude Phase A Voltage Phase
1 1 0.955
True True
SE 0.5 SE 0.95 114
0 5 10 0 5 10
Bus number Bus number
0

−0.5
Voltage Phase (degree)
Voltage Magnitude (pu)

0.95 Fig. 6. Phase C estimation results and true value

−1

−1.5
random error. Estimation results for phase B and C are also
−2
0.9
shown in Figure 5, 6.
−2.5 So far, only five 𝜇PMUs have been used in the simulation
−3 since this is the least number that can make the system
observable. In order to analysis the effect of redundancy on
−3.5
estimation results, another 𝜇PMU was assumed to be installed
0.85 −4 at bus 2. Therefore, voltage phasor at bus 2 was measured,
0 5 10 0 5 10
Bus number Bus number so as the current phasor of all incident lines. Note that there
are distributed loads on the line connected bus 2 and bus 7,
Fig. 4. Phase A estimation results and true value so current leaves bus 2 to bus 7 was not measured. Hence,
the sixth 𝜇PMU brought 3 more voltage measurements and
8 more current measurements. The state estimator gave more
In Figure 4, it can be seen that the estimation results are accurate results with this six 𝜇PMUs. Estimation results of
quite accurate even though measurements are biased with voltage magnitude of phase C is shown in figure 7. From
figure 7, the state estimation results has been improved greatly [3] M. Gol and A. Abur, “A robust pmu based three-phase state estimator
by additional measurements. using modal decoupling,” Power Systems, IEEE Transactions on, vol. 29,
no. 5, pp. 2292–2299, 2014.
[4] ——, “Synchro-phasor based three phase state estimation using modal
Phase C Voltage Magnitude components,” in Innovative Smart Grid Technologies (ISGT Europe),
1 2012 3rd IEEE PES International Conference and Exhibition on. IEEE,
True
SE with 5 microPMU
2012, pp. 1–4.
0.995
SE with 6 microPMU [5] D. A. Haughton and G. T. Heydt, “A linear state estimation formulation
0.99 for smart distribution systems,” IEEE Transactions on Power Systems,
no. 28, pp. 1187–1195, 2013.
0.985 [6] S. Sarri, M. Paolone, R. Cherkaoui, A. Borghetti, F. Napolitano, and
Voltage Magnitude (pu)

C. A. Nucci, “State estimation of active distribution networks: compari-

0.98
son between wls and iterated kalman-filter algorithm integrating pmus,”
0.975 in Innovative Smart Grid Technologies (ISGT Europe), 2012 3rd IEEE
PES International Conference and Exhibition on. IEEE, 2012, pp. 1–8.
0.97 [7] M. Pau, P. Pegoraro, and S. Sulis, “Efficient branch-current-based distri-
bution system state estimation including synchronized measurements,”
0.965 Instrumentation and Measurement, IEEE Transactions on, vol. 62, no. 9,
0.96
pp. 2419–2429, 2013.
[8] A. von Meier, D. Culler, A. McEachern, and R. Arghandeh, “Micro-
0.955 synchrophasors for distribution systems,” in IEEE 5th Innovative Smart
Grid Technologies Conference, Washington, DC, 2014.
0.95 [9] M. Wache and D. Murray, “Application of synchrophasor measurements
0 2 4 6 8 10 12 14
Bus number for distribution networks,” in Power and Energy Society General Meet-
ing, 2011 IEEE. IEEE, 2011, pp. 1–4.
[10] H. Wang and N. N. Schulz, “A revised branch current-based distribution
Fig. 7. Phase C estimation results and true value system state estimation algorithm and meter placement impact,” Power
Systems, IEEE Transactions on, vol. 19, no. 1, pp. 207–213, 2004.
[11] M. Gol, “Incorporation of pmus in power system state estimation,” Ph.D.
dissertation, Northeastern University Boston, 2014.
V. C ONCLUSION [12] M. Zhou, V. Centeno, J. S. Thorp, A. G. Phadke et al., “An alternative
for including phasor measurements in state estimators,” Power Systems,
A linear three phase state estimator is applied to distribution IEEE Transactions on, vol. 21, no. 4, pp. 1930–1937, 2006.
system in this paper. The estimator uses a full three-phase [13] K. Das, J. Hazra, D. P. Seetharam, R. K. Reddi, and A. K. Sinha,
“Real-time hybrid state estimation incorporating scada and pmu mea-
representation of distribution network, so that it works well surements,” in Innovative Smart Grid Technologies (ISGT Europe), 2012
even the network is three phase unbalanced. According to the 3rd IEEE PES International Conference and Exhibition on. IEEE, 2012,
simulation results, it is found the estimator is able to give three pp. 1–8.
[14] A. Abur and A. G. Exposito, Power system state estimation: theory and
phase estimation with good accuracy. Measurements used in implementation. CRC Press, 2004.
the estimator are all synchronized phasors which makes the [15] B. Alcaide-Moreno, C. Fuerte-Esquivel, M. Glavic, and T. Van Cutsem,
estimator linear and the solution non-iterative and can be “Hybrid processing of scada and synchronized phasor measurements for
tracking network state,” in IEEE PES General Meeting.
obtained by 𝜇PMU in reality. Considering high sampling rate [16] K. D. Jones, J. S. Thorp, and R. M. Gardner, “Three-phase linear state
and high precision measurements of 𝜇PMU, the estimator is estimation using phasor measurements,” in Power and Energy Society
potential to realize real time monitoring in distribution system. General Meeting (PES), 2013 IEEE. IEEE, 2013, pp. 1–5.
[17] J. Liu, J. Tang, F. Ponci, A. Monti, C. Muscas, and P. A. Pegoraro,
Future work could include comparison of decoupled state “Trade-offs in pmu deployment for state estimation in active distribution
estimator and phase state estimator. Theoretically decoupled grids,” Smart Grid, IEEE Transactions on, vol. 3, no. 2, pp. 915–924,
state estimator can decrease the computation time if parallel 2012.
[18] M. Ahmad, Power system state estimation. Artech House, 2013.
processor is available. However, the decoupled state estimator [19] W. H. Kersting, “Radial distribution test feeders,” in Power Engineering
is not as accurate as phase estimator because the admittance of Society Winter Meeting, 2001. IEEE, vol. 2. IEEE, 2001, pp. 908–912.
transmission line is strictly symmetrical. It is also of interest to [20] P. S. Limited, “Pqube 3 technical specifications,” http://www.
powersensorsltd.com/PQube3.php#technical.
include existing SCADA measurements in the state estimator [21] K. D. Jones, “Three-phase linear state estimation with phasor mea-
to improve redundancy. surements,” Ph.D. dissertation, Virginia Polytechnic Institute & State
University, 2011.
ACKNOWLEDGMENT
This project is funded by the National Research Foundation
Singapore under its Campus for Research Excellence and
Technological Enterprise (CREATE) programme.
R EFERENCES
[1] F. Ding and C. D. Booth, “Applications of pmus in power distribution
networks with distributed generation,” in Universities’ Power Engineer-
ing Conference (UPEC), Proceedings of 2011 46th International. VDE,
2011, pp. 1–5.
[2] A. E. Kahunzire and K. O. Awodele, “Improving distribution network
state estimation by means of phasor measurement units,” in Power
Engineering Conference (UPEC), 2014 49th International Universities.
IEEE, 2014, pp. 1–6.