Utility Customer Model for Fifth Harmonic State
Estimation in Combination with Fundamental State
Estimation in Distribution Network
Naotaka Okada, Member, IEEE
System Engineering Research Laboratory
Central Research Institute of Electric Power Industry
Tokyo, Japan
naotaka@ieee.org
Abstract— In the last decade, harmonic analysis of Japanese
middle voltage (MV) customer and low voltage (LV) customer
has been done, the distribution of the fifth harmonic current can
be classified MV customer’s single phase load, LV customer’s
three phase and single phase load named group A and MV three
phase load named group B, and the amplitude per load kW has
been investigated from sum of each current vector. By the
application of this knowledge, there is a possibility to estimate
the harmonic distribution with many harmonic source in the
distribution network. In this paper, we demonstrate a harmonic
state estimation (HSE) in combination with fundamental state
estimation (FSE) method with limited measurement condition
by using the fifth harmonic current model of utility customer.
Numerical experiments are conducted in seven node distribution
network with transient simulation to test the proposed method.
The results show that the proposed HSE in combination with
FSE by using utility customer model approach can reliably
identify harmonic distribution with distributed load conditions,
background harmonic voltage and shunt capacitors.
Index Terms—Harmonic state estimation, fifth harmonic,
harmonic phase angle, harmonic current model, distribution
network, fundamental state estimation.
I. INTRODUCTION
By the spray of power electronics technology, harmonic
situation has changed to be generated from many utility
customer. It is difficult to measure all utility customer in a
distribution feeder at same time. Under the terms of low
number of measurement, it is difficult to determine effectual
harmonic current combination to conform calculation closely
to measurement.
State estimation technique of power systems was
originally developed for transmission networks [1]-[2]. These
techniques have been adapted for distribution networks [3].
And three phase state estimation algorithm were developed
for distribution networks [4]-[5]. Impact of measurement
location on accuracy was demonstrated for utility service area
978-1-4673-8040-9/15/$31.00 ©2015 IEEE
Kenji Yukihira
System Engineering Research Laboratory
Central Research Institute of Electric Power Industry
Tokyo, Japan
model [6] and 95-bus test model [7]. Extended Kalman Filter
technique [8] has been tested for distribution networks. And,
the bad data detection method has been mainly discussed for
power transmission networks [9]-[11].
Harmonic state estimation (HSE) is used to estimate
harmonic distributions in power transmission networks [12],
[13]. This will enable to locate harmonic sources. Kalman
Filter technique has been also applied to HSE [14]-[16]. As
basic HSE method, the weighted least squares (WLS) based
method has been developed [17] and tested [18] in a
transmission network in Japan. Here eight synchronous
meters are used to estimate seven harmonic current sources.
Due to high expense of harmonic measurement system,
only limited number of harmonic meter are available in
power transmission network. To solve this problem, optimal
meter placement method has been developed [19], [20].
However, in case of the number of meter less than the
harmonic sources, these HSE cannot estimate harmonic
sources. On the other hand, sparsity maximization method
has been proposed to solve this estimation problem supposing
spatial sparsity of harmonic current in power transmission
network [21]. In case of distribution network with many
harmonic source, there is possibility spatial evenness of
harmonic current.
In this paper, we demonstrate HSE in combination with
fundamental state estimation (FSE) method with limited
measurement condition by using the fifth harmonic current
model of utility customer. And numerical experiments are
conducted in seven node distribution network with transient
simulation to test the proposed method. The results show that
the proposed HSE in combination with FSE by using
proposed utility customer model approach can reliably
identify harmonic distribution with distributed load
conditions, back ground harmonic voltage and shunt
capacitors.
 II. CHARACTERISTICS OF FIFTH HARMONIC CURRENT ON
UTILITY CUSTOMER IN JAPAN
The fifth harmonic mainly generate failure. Therefore, it is
important to understand this mechanism. The fifth harmonic
current of 81 single phase transformer and 87 three phase
transformer in middle voltage (MV) utility customer have
been analyzed in reference [22]. And also the fifth harmonic
current of 30 single phase transformer in low voltage (LV)
utility customer have been analyzed in reference [23].
A. Phase Angle
JIS C 4304-2005 “6 kV Oil-immersed distribution
transformers” (Japanese Industrial Standards) specify the
three phase transformer of the capacity from 75kVA to
500kVA Y-delta connection for primary-secondary.
Therefore, the three phase load of the MV customer supplied
from this Y-delta transformer. On the other hand, single
phase loads and LV three phase loads were supplied form
line-to-line single phase transformer and delta-delta
transformer, as shown in figure 1.
MV line
Group A
MV customer
Single phase load
LV customer
Three phase load
Single phase load
Group B
MV customer
Three phase load
In MV utility customer, the RMS current per load kW has
been investigated from sum of each current vector. RMS of
fifth harmonic current per 1kW of load are 3.5 ~ 7.5mA/kW
for single phase transformer and 5 ~ 11mA/kW for three
phase transformer [22].
In LV utility customer, RMS of fifth harmonic current per
1kW of load are 4 ~ 8mA/kW for single phase load and
27.1mA/kW for three phase load [23]. The RMS harmonic
current of three phase load in LV customer was 4.5 times for
single phase load. This result may be caused by low load
level for only three phase load of LV customer, in this
investigation. An investigation of harmonics attenuation and
diversity among distributed single-phase power electronic
loads is shown in reference [24].
III. UTILITY CUSTOMER MODEL FOR STATE ESTIMATION
The WLS method is a powerful tool to determine most
probably system state from measurement data with random
noise. We use the WLS method for state estimation of
distribution network. Fundamental and fifth harmonic model
of utility customer are described here.
A. Power and Reactive Power Model of Utility Customer
To compensate for the lack of measurement data, pseudo
measurement data has been used in state estimation of
distribution system. The pseudo measurement data of
customer load is based on customer monthly kWh, kW rating,
load type and so on. In this paper, we try to estimate
fundamental phase angle of nodes in a way that sets the
pseudo measurement data based on a simple assumption.

N P ,V 2 
Figure 1. Transformer winding for single phase, LV three phase and MV
three phase loads in Japan.
2V
For these reason, the distribution of fifth harmonic current
vector in MV line, as shown in figure 2, can be classified MV
customer’s single phase load, LV customer’s three phase and P  2V P P  2V
single phase load named group A (supplied by line-to-line
single and delta-delta transformer) and MV three phase load Figure 3. Normal distribution model for State Estimation.
named group B (supplied by Y-delta transformer) [22], [23].
As shown in figure 3, we assume a normal distribution of
90㼻
average demand μPi and standard deviation σPi as demand
Phase angle data for each node i in the distribution feeder. The μPi is used
Group A as measurement data for the lack of demand measurement
MV customer 180㼻 0㼻
Single phase load Group B
data. And the σPi is set as weight of the WLS method.
LV customer MV customer Concretely, average power μPi and standard deviation σPi
Three phase load Three phase load
Single phase load of node i is given by the following.
P max i  P min i  P max i  P min i  
270㼻
 PPi , VPi 
Figure 2. Distribution of two fifth harmonic current vector groups on utility 2 4
customers.
Where, Pmaxi set based on maximum or contract capacity of
customer for each harmonic current group at node i. Similarly,
B. Amplitude Pmini set zero or minimum capacity of customer for each
Root mean square (RMS) currents of the fifth harmonic harmonic current group at node i. Average reactive power μQi
converted MV distribution line phase current are discussed and standard deviation σQi of node i are obtained as multiply
here. (1) by power factor of utility customers.
B. Fifth Harmonic Current Model of Utility Customer In convergence calculation, under the condition of
Normal distribution model, shown in figure 3, apply to wJ / wx 0 from equation (4), state variable xk update
fifth harmonic current phase angle and RMS of amplitude. following equation.
The fifth harmonic current angle characteristic at node i
x k  Px k H x k  R 1 >z  hx k @
T
utility customer in the distribution network modeled average x k 1
  
whx 
μθ5i and standard deviation σθ5i by reference to [22] and [23].
Similarly, the RMS of harmonic current amplitude
P x  >H x T
R 1 H x @ 1
, H x 
wx
modeled average μA5i and standard deviation σA5i.
B. Fundamental Measurement Equation
 PA5i K 5i ˜ PPi , VA5i K 5i ˜ VPi   For the node voltage measurement of the fundamental, the
measurement equation h(x) for real and imaginary part of
Where, K5i is harmonic current per load kW.
fundamental voltage Vi at node i are

IV. HSE IN COMBINATION WITH FSE
By the component of line inductive reactance, when
increase of load flow, the voltage phase angle of fundamental
at end of feeder lag to norm bus voltage. Because of non-
linear load generate harmonic current for voltage at
connection node, fifth harmonic current shift 5 times for
fundamental voltage angle shifting in fifth harmonic space.
 ReVi  Ei , ImVi  Fi  
, where, Ei and Fi are real and imaginary part of node i
voltage Vi and indicate state variables x. The measurement
noise assume 2σVzi as maximum error guaranteed by
measurement Vzi. σVzi are set on diag(R).
For the polar coordination, h(x) for the amplitude and
phase angle of node i voltage are following.
 §F ·
arctan¨¨ i ¸¸  
Measurement Network and
data customer data Vi Ei2  Fi 2 , Ti
Power and reactive Fundamental state estimation
Network model © Ei ¹
power data of feeder (FSE) Power and reactive
power model of The measurement equations h(x) of power Pi and reactive
Fundamental voltage
utility customers
power Qi on utility customers at node i are shown in
phase angle
following equations.

¦ E G E j  E i Bij F j  Fi Gij F j  Fi Bij E j 

Harmonic voltage Harmonics state estimation Harmonic current N
and current data of
(HSE)
model of utility
Pi i ij
  
feeder customers
j 1

¦ F G E j  Ei Bij E j  E i Gij F j  Fi Bij F j 

N
Qi i ij
Distribution of harmonic voltage and current j 1

Where, Gij and Bij are element of fundamental nodal

Figure 4. Calucation flow of harmonic distribution by HSE in combiation
with FSE. admittance matrix Y. Following form were used in this paper.
 Yij Gij  jBij  
This phase shift of fundamental voltage make influence for
estimation result of harmonic when unconsidered this phase The measurement equations h(x) of line power flow Pij and
shift. HSE in combination with FSE method are described Qij from node i to node j are shown in following equations.
here. Figure 4 shows calculation flow of HSE in combination
with FSE. Pij E i Gij E j  E i Bij F j  Fi Gij F j  Fi Bij E j
A. Estimation by the WLS   E ii2 Gij  Fii2 Gij  
Generally, the relation with measurement z, power system Qi Fi Gij E j  E i Bij E j  E i Gij F j  Fi Bij F j
or measurement equation h(x), state variables x and random
noise n are shown as the following.  E ii2 Bij  Fii2 Bij

 z hx   n 
 C. Fifth Harmonic Measurements Equation
For the node voltage measurement, the measurement
Estimated state variables xe are given by the state x which
equation h(x) for real E5i and imaginary F5i part of fifth
minimize the following equation J.
harmonic voltage V5i at node i are obtained by adding suffix
 J >z  hx @ R 1 >z  hx @   5 to (6). The E5i and F5i indicate state variables x. The
T

measurement noise assume 2σVz5i as maximum error

R is noise covariance matrix. Here, the diagonal elements Rii guaranteed by measurement Vz5i.
set to σi2. The σi is standard deviation of noise for The measurement equation h(x) for RMS of harmonic
measurement zi. For example, we assumed deviation 2σi as current at node i is
maximum error guaranteed by measurement zi.
 Figure 5 shows a simple overhead 6.6kV distribution network
I 5i C 5 i2  D5 i2
model which feeds six areas by one feeder. And the feeder
  
¦ G5 E 5 j  B5 ij F 5 j 
N
operation and maximum capacity are assumed 3MVA and
C5i ij
j 1 4MVA, respectively.
V6 6 B Group A: Low voltage load

¦ G5 ij F 5 j  B5 ij E 5 j 
N
I56 Group B: Middle voltage three phase load
D5 i Measurement
j 1 Upper SC with inductor
Vz1,
V50 Vz51 V5
, where, G5ij and B5ij are element of fifth harmonic nodal V2 V3
B
V4
A A
0 1 2 3 4 5
admittance matrix Y5. Form adding suffix 5 to (9) was used I53 I54 I55

in this paper. I52

V7
The measurement equation h(x) for phase angle of fifth Pz12 , Qz12
Iz512
A
7 B
SC without inductor I57
harmonic current at node i is Measurement

 D5 i  
Distribution substation
‘I 5i arctan SC without inductor
C 5i Group A model

For the line current measurement from node i to node j, the Group B model

measurement equation h(x) for real and imaginary part of LR load Diode rectifier

LR load and diode rectifier

fifth harmonic current I5ij are

ReI 5ij  G5ij E 5 j  B5ij F 5 j  G5ij E 5i  B5ij F 5i
LR load and diode rectifier
LR load Diode rectifier

  
ImI 5ij   B5ij E 5 j  G5ij F 5 j  B5ij E 5i  G5ij F 5i Figure 5. A distribution network model.

The measurement noise assume 2σIz5ij as maximum error Voltage measurement Vz1 and Vz51 are obtained at
guaranteed by measurement Iz5ij. secondary bus of distribution substation transformer. Line
D. Fifth Harmonic Node Admittance Matrix power flow and current measurement Pz12, Qz12, and Iz512 are
The formation of fifth harmonic node admittance matrix obtained at feeder from the bus.
Y5 is needed to calculate harmonic measurement equation. B. Condition of Utility Customer
This formation realize 5 times for inductive reactance and 1/5 In this simulation model, group A customer assume node 2,
times for capacitive reactance based on branch information of 4 and 5. And group B customer assume node 3, 6 and 7.
fundamental component. To compensate reactive power of load, shunt capacitor
E. Consideration of Fundamental Voltage Phase Angles (SC) are installed mainly MV customer in Japanese
To consider fundamental voltage phase angle, average distribution network. SC of MV customer were set at node 3,
phase angle μθ5i of harmonic current model is corrected by 6 and 7.
following equation. Line impedances of the distribution network,
characteristics of fifth harmonic current of group A and group
 PT 5ci PT 5i  5 ˜ 'T i   B, setting of customer capacity, and SC capacity of MV
customer are set up by reference to [26].
where μθ5’i is corrected phase angle of harmonic current
model at utility customer and Δθi is phase difference of C. Hramonics load model for Transient Simulation
fundamental voltage at node i for norm voltage angle. Harmonic loads were modeled by diode rectifier and
F. Formation of Estimation Model parallel resistor-inductor (RL) load. For group A, as shown in
figure 5 single phase diode rectifier and parallel RL load were
Formation of z, h(x), diag(R) (diagonal elements of matrix
placed each phase to phase by single phase transformer. For
R) and x are following. The average μ is set in the
group B, three phase diode rectifier and parallel RL load was
measurement vector z with real measurement value. The
connected by Y-delta transformer.
measurement equations are set in the h(x). The standard
deviation σ is set in the diag(R) with the measurement error. D. Result of FSE and Fifth HSE
The node voltage is set in the state variable x. For true value calculation, numerical experiments are
V. VALIDATION BY SIMULATION conducted by using transient simulation with fifth harmonic
voltage distortion of upper back network voltage distortion of
Numerical experiments are conducted in seven node 1.0% and phase angle of 0 degree and distributed customer
distribution network model to test the proposed method by load condition. After true value calculation, FSE and HSE are
using transient simulation with MATLAB Simulink. conducted by using Vz1, Pz12, Qz12, Vz51, and Iz51 as
A. A Distribution Network Model measurement value.
Medium voltage distribution feeders are mainly Figure 6 shows comparison of true with fundamental
constructed by overhead 6.6kV distribution network, in Japan. voltage of FSE result.
 1.2 䢳䢰䢴 [6] Ke Li, "State estimation for power distribution system and
Fundamental voltage

True measurement impacts," IEEE Trans. Power Systems, vol. 11, No. 2, pp.
䢳䢰䢳
Estimation 911 - 916, May 1996.
rms (pu)

[7] R. Singh, B.C. Pal, R.B. Vinter, “Measurement Placement in

ᇶᮏἼ㟁ᅽᐇຠ್䢢䣘

1.0 䢳
Distribution System State Estimation”, IEEE Transactions on Power
Systems, Vol. 24, No. 2, 2009, pp. 668 - 675.
[8] R. Gelagaev, P. Vermeyen, J. Driesen, “State estimation in distribution
䢲䢰䢻

grids”, ICHQP 2008. 13th International Conference on Harmonics and

0.8 䢲䢰䢺

Quality of Power, 2008, pp. 1 – 6.

䢳 䢴 䢵 䢶 䢷 䢸 䢹

10 䢳䢲

[9] E. Handschin, F.C. Schweppe, J. Kohlas, A. Fiechter, “Bad data

Fundamental voltage
phase angle (deg.)

䢷 analysis for power system state estimation”, IEEE Transactions on

Power Apparatus and Systems, Vol. 94 , No. 2 , 1975 , pp. 329 - 337.
ᇶᮏἼ㟁ᅽ఩┦䢢䣦䣧䣩

0 䢲
[10] R. Baldick, K.A. Clements, Z. Pinjo-Dzigal, P.W. Davis,
“Implementing nonquadratic objective functions for state estimation
and bad data rejection”, IEEE Transactions on Power Systems, Vol. 12,
䢯䢷

-10 䢯䢳䢲
No. 1, 1997, pp. 376 - 382.
[11] I.W. Slutsker, “Bad data identification in power system state estimation
䢳 䢴 䢵 䢶 䢷 䢸 䢹

Vz1 V2 V3 V4 V5 V6 V7
based on measurement compensation and linear residual calculation”,
Figure 6. Comparison of true with fundamental voltage of FSE. IEEE Transactions on Power Systems, Vol. 4, No.1, 1989, pp. 53 - 60.
[12] G.T. Heydt, "Identification of harmonic sources by a state estimation
technique", IEEE Transactions on Power Delivery, Vol. 4, No. 1, 1989,
Figure 7 shows comparison of true with fifth harmonic of pp. 569-576.
HSE result. [13] A.P.S. Meliopoulos, Fan Zhang, S. Zelingher, "Power system harmonic
state estimation", IEEE Transactions on Power Delivery, Vol. 9, No. 3,
20 䢴䢲
1994, pp. 1701-1709.
Measurement at feeder
Fifth voltage/10 rms (V),

Harmonic voltage

䢳䢺
Harmonic current of [14] H.M. Beides, G.T. Heydt, "Dynamic state estimation of power system
voltage and current
Fifth current rms (A)

䢳䢸
customers harmonics using Kalman filter methodology", IEEE Transactions on
㧗ㄪἼ㟁ᅽ䢱䢳䢲䢢䣘䢮䢢㟁ὶ䢢䣃凚ᐇຠ್凛

of upper bus

䢳䢶

䢳䢴
Power Delivery, Vol. 6, No. 4, 1991, pp. 1663-1670.
10 True
䢳䢲

䢺 Estimation [15] K.K.C. Yu, N.R. Watson, J. Arrillaga, "An adaptive Kalman filter for
䢸 dynamic harmonic state estimation and harmonic injection tracking",
䢶
IEEE Transactions on Power Delivery, Vol. 20, No. 2, 2005, pp. 1577-
0
䢴

䢲
1584.
[16] H. Ma, A.A. Girgis, "Identification and tracking of harmonic sources in
䢳 䢴 䢵 䢶 䢷 䢸 䢹 䢺 䢻

180 䢳䢺䢲

Group A a power system using a Kalman filter", IEEE Transactions on Power

angle (deg.)

䢳䢵䢷

䢻䢲
Delivery, Vol. 11, No. 3, 1996, pp. 1659-1665.
Fifth harmonics

䢶䢷
[17] K. Yukihira, “Development of a Program for Estimating Harmonic
phase 㧗ㄪἼ఩┦䢢䣦䣧䣩

0 䢲

Current Sources”, CRIEPI Report, T89043, May 1999.

䢯䢶䢷

䢯䢻䢲
Measurement at feeder
䢯䢳䢵䢷
voltage and current
-180 䢯䢳䢺䢲
䢳 䢴
V50 Vz51 Iz512 I52
Figure 7. Comparison of true with fifth harmonic of HSE.
The results show that the proposed approach can reliably
identify fundamental voltage and fifth harmonic distribution.
VI.
In this paper, we demonstrate that the proposed utility
customer model for fifth HSE in combination with FSE in
distribution network approach can reliably identify harmonic
distribution by using transient simulation.
[1] R. E. Larson, W. F. Tinney and J. Peschon, "State Estimation in Power
Systems Part I-II," IEEE Trans. Power Apparatus and Systems, vol.
PAS-89, No. 3, pp. 345-363, Mar. 1970.
[2] F. C. Schweppe, J. Wildes and D. B. Rom, "Power System Static-State
Estimation Part I-III," IEEE Trans. Power Apparatus and Systems, vol.
PAS-89, No. 1, pp. 120-135, Jan. 1970.
[3] I. Roytelman, S.M. Shahidehpour, “State estimation for electric power
distribution systems in quasi real-time conditions”, IEEE Transactions
on Power Delivery, Vol. 8, No. 4, 1993, pp. 2009 - 2015.
[4] M.E. Baran, A.W.Kelley, “State estimation for real-time monitoring of
distribution systems”, IEEE Transactions on Power Systems, Vol. 9,
No. 3, 1994, pp. 1601 - 1609.
[5] C.N. Lu, J.H. Teng, W.-H.E. Liu, “Distribution system state
estimation”, IEEE Transactions on Power Systems, Vol. 10, No. 1,
1995, pp. 229 – 240.
[18] N. Kanao, M. Yamashita, H. Yanagida, M. Mizukami, Y. Hayashi, J.
Group B Matsuki, "Power system harmonic analysis using state-estimation
䢵 䢶 䢷 䢸 䢹 䢺 䢻
method for Japanese field data",IEEE Transactions on Power Delivery,
I53 I54 I55 I56 I57 Vol. 20, No. 2, 2005, pp. 970-977.
[19] J.E. Farach, W.M. Grady, A. Arapostathis, "An optimal procedure for
placing sensors and estimating the locations of harmonic sources in
power systems", IEEE Transactions on Power Delivery, Vol. 8, No. 3,
1993, pp. 1303-1310.
[20] A. Kumar, B. Das, Jaydev Sharma, "Genetic algorithm-based meter
placement for static estimation of harmonic sources", IEEE
Transactions on Power Delivery, Vol. 20, No. 2, Part 1, 2005, pp. 1088-
CONCLUSIONS 1096.
[21] Huaiwei Liao, "Power System Harmonic State Estimation and
Observability Analysis via Sparsity Maximization", IEEE Transactions
on Power Systems, Vol. 22, No. 1, 2007, pp. 15-23.
[22] K. Yukihira, N. Okada, “The Suppressing Effect of Voltage Distortions
in Electric Power Systems by Harmonic Currents Emitted by
Customers with High Voltage Power Supply”, CRIEPI Report, R11003,
REFERENCES Oct. 2011.
[23] K. Yukihira, N. Okada, K. Ishibashi, “Characteristics of harmonic
currents emitted by customers with low-voltage power supplies -
Analysis of data measured in the metropolitan area in winter heavy-
load season -”, CRIEPI Report, R13004, Feb. 2014.
[24] A. Mansoor, W.M. Grady, A.H. Chowdhury, M.J. Samotyi, "An
investigation of harmonics attenuation and diversity among distributed
single-phase power electronic loads", IEEE Transactions on Power
Delivery, Vol. 10, Issue 1, 1995, pp. 467-473.
[25] N. Okada, M. Takasaki, K. Tanaka, "A case study of bad data detection
for distribution feeder voltage measurement", ISGT Europe, 2012. pp.
1-6.
[26] N. Okada, K. Yukihira, “Harmonic State Estimation in Distribution
Network Based on Fifth Harmonic Current Characteristic”, ICHQP
2014, The 16th International Conference on Harmonics and Quality of
Power.