Energy and Power Engineering, 2020, 12, 73-83

https://www.scirp.org/journal/epe
ISSN Online: 1947-3818
ISSN Print: 1949-243X

State Estimation of Distribution Network

Considering Data Compatibility

Shengtao Wu, Yan Li

School of Electrical and Electronic Engineering, Huazhong University of Science and Technology, Wuhan, China

How to cite this paper: Wu, S.T. and Li, Y. Abstract

(2020) State Estimation of Distribution
Network Considering Data Compatibility. Considering that the measurement devices of the distribution network are
Energy and Power Engineering, 12, 73-83. becoming more and more abundant, on the basis of the traditional Supervi-
https://doi.org/10.4236/epe.2020.124B008
sory Control And Data Acquisition (SCADA) measurement system, Phasor
Received: March 12, 2020
measurement unit (PMU) devices are also gradually applied to the distribu-
Accepted: April 7, 2020 tion network. So when estimating the state of the distribution network, the
Published: April 10, 2020 above two devices need to be used. However, because the data of different
measurement systems are different, it is necessary to balance this difference
so that the data of different systems can be compatible to achieve the purpose
of effective utilization of the estimated power distribution state. To this end,
this paper starts with three aspects of data accuracy of the two measurement
systems, data time section and data refresh frequency to eliminate the differ-
ences between system data, and then considers the actual situation of the
three-phase asymmetry of the distribution network. The three-phase state es-
timation equations are constructed by the branch current method, and finally
the state estimation results are solved by the weighted least square method.

Keywords
Distribution Network, State Estimation, Data Compatibility, Branch Current
Method

1. Introduction
The state estimation algorithm mainly improves the accuracy of the state estima-
tion by increasing the redundancy of the data, so the state estimation is also
called filtering [1]. Traditional state estimation is mainly used in transmission
grids, and there are relatively few applications in distribution networks. Until
recent years, the state estimation of distribution networks has been widely con-
cerned by researchers. However, the distribution network is different from the

DOI: 10.4236/epe.2020.124B008 Apr. 10, 2020 73 Energy and Power Engineering

S. T. Wu, Y. Li

transmission network and has its own characteristics, such as three-phase asym-
metry. Therefore, the state estimation of the transmission system is not suitable for
the distribution network [2]. Therefore, it is necessary for us to conduct targeted
research on the state estimation of the power distribution system.
With the development of distribution automation technology, measurement
devices for distribution networks have gradually increased, and monitoring of
distribution systems has become more and more detailed. The types of distribu-
tion network measurement devices are also getting richer. On the basis of the
SCADA measurement system, PMU devices have also appeared in the distribu-
tion network, and the data accuracy has improved. In reference [3], the author
proposed a state estimation method for distribution network systems with
high-precision PMU measurement data, but the establishment of the method did
not specify the specific details of its construction. In reference [4], the author al-
so highlights the role of high-precision PMU data for state estimation, and ana-
lyzes the advantages and disadvantages of state estimation in polar and rectan-
gular coordinate systems. The results show that the accuracy of distribution
network state estimation is improved. In general, it is better to choose a rectan-
gular coordinate system for solving. In reference [5], the author believes that
there will be a PMU measurement system in the distribution network in the fu-
ture. Using this high-precision system, the error of the state variables is mini-
mized by optimizing the scheme to achieve the purpose of improving the esti-
mation accuracy.
However, there are many differences in the data of these two systems. Gener-
ally, the SCADA data has insufficient accuracy, no time scale, and the data re-
fresh frequency is low. The PMU data has high accuracy, but the number of in-
stallations in the distribution network is relatively small but the data frequency is
refreshed rate is higher. If it is not processed, the effect of the distribution net-
work state estimation will be greatly reduced. However, there is relatively little
research in this area. In reference [6], the author considers the problem of in-
consistent time sections in the transmission process of different measuring de-
vices. Therefore, the time delay error is used to compensate for the error caused
by the communication delay, and the problem caused by the inconsistent time
section is improved. The problems brought by it ultimately improve the accura-
cy of state estimation. In reference [7], the author used the curve fitting method
to make up for the data of the measurement system with low data refresh fre-
quency, but the data accuracy needs to be improved. In reference [8], the author
mainly considered the impact of PMU data on the accuracy of state estimation,
but did not mention the data difference between different systems. In reference
[9], the author considered the fusion of PMU data and SCADA data, and used a
hybrid state estimation algorithm to solve it, but for the problem of inconsistent
time sections, no effective solution was proposed.
Based on the above research, this article first analyzes the different characte-
ristics of SCADA and PMU data, and then based on the specific differences be-

DOI: 10.4236/epe.2020.124B008 74 Energy and Power Engineering

 S. T. Wu, Y. Li

tween them, gives corresponding solutions from three aspects: data accuracy,
time section, and data refresh frequency. Then, based on the three-phase line
model of the distribution network, the branch current method is used to esti-
mate the state of the distribution network considering data compatibility.

2. Difference Analysis between SCADA Data and PMU Data

So far, medium voltage distribution feeders lack the necessary infrastructure to
ensure the observability of the network [10]. Considering the economics of
power grid deployment facilities, multiple measurement systems need to coope-
rate to provide measurement data. The two main devices in the system are in-
troduced and compared below.
The PMU measures the DC, switching, phase-detection pulse, terminal vol-
tage, terminal current, and three-phase current, three-phase voltage, and
switching signals of the unit in real time. Then, the PMU calculates the meas-
ured signals to obtain the internal potential, speed, and electrical quantities such
as active power, reactive power, and frequency of the line, and can perform pri-
mary / secondary value conversion. The real-time output rate of the PMU is in
the millisecond level, and some literatures specifically record that data is trans-
mitted once every 10 ms, which can be selected according to the master station.
SCADA systems collect data at HV-MV substations, typically at rates ranging
from a few seconds to a minute. For radial feeders, this measurement usually in-
cludes the medium-voltage bus voltage amplitude and bus current, assuming the
load has an average power factor, so that the total P and total Q delivered by the
feeder can be calculated. So far, this is actually the only telemetry information at
the medium voltage level for most utilities [11].
In Table 1, it compares these two measurements. It can be seen that the PMU
data has higher accuracy and time stamping, while the SCADA data has lower
accuracy and no time stamping, and the refresh frequency is lower than that of
PMU. Therefore, when performing state estimation, a compatibility analysis of
these two types of data is required.

3. Analysis of Two Types of Data Compatibility

3.1. Analysis of Data Precision Compatible Methods
Since the accuracy of the three-phase data of the SCADA system is lower than that
of the three-phase data of the PMU system, in a medium-voltage distribution

Table 1. Comparison of SCADA and PMU data.

Measurement
Measurement data Refresh rate Accuracy class Time stamp
system

Three-phase voltage, three-phase

SCADA Second level Lower No
current, three-phase power

Three-phase voltage phasor, Millisecond

PMU high Yes
three-phase current phasor level

DOI: 10.4236/epe.2020.124B008 75 Energy and Power Engineering

S. T. Wu, Y. Li

network that mixes two types of system data, if they cannot be distinguished
well, it may be an improvement for the medium-voltage distribution network
The accuracy of state estimation will not help much. At present, a common me-
thod is to use the error distribution of each system as the weight of the system
data. Then high-precision data will naturally have a high weight coefficient, and
low-precision data will have a lower weight. This article will not go into too
much detail.

3.2. Analysis of Data Time Section Compatibility Methods

Because the three-phase data of the SCADA system does not have a time stamp,
the three-phase data of the PMU system has a time stamp. Then, in the state es-
timation of the distribution network, because the time scale of the SCADA
three-phase data cannot be accurately understood, this will cause the state esti-
mation to be significantly different from the actual state of the system, and
therefore will have a greater impact on the accuracy of the state estimation.
Based on this, this article considers that the upload frequency of the PMU is
milliseconds, and assumes that the specific upload rate is set to 10 ms. Under
normal circumstances, this article believes that within a 10 ms time interval, the
possibility of a large change in the related electricity parameters is small or even
negligible. Therefore, in order to add timestamps to the relevant SCADA
three-phase data, and also to ensure that the SCADA three-phase data and the
PMU three-phase data can be at the same time section, this article will introduce
the introduction of buffers in the computer and define the concept of time buf-
fers. That is, by setting a time buffer with a time interval of 10 ms at the time of
PMU three-phase data upload, and thinking that as long as there is relevant
SCADA three-phase data upload in this time zone, the time scale of the SCADA
three-phase data is related to PMUs have the same time scale. This not only adds
time scales to the SCADA three-phase data, but also unifies the PMU three-phase
data and SCADA three-phase data on the same time section, which facilitates the
subsequent state estimation of the distribution network.
As shown in Figure 1, it is assumed that the three-phase data of the PMU is
uploaded at 10 milliseconds. At this time, the three-phase data of the SCADA is
not uploaded, but is slightly later than the PMU, and also in the time buffer

Figure 1. Adding SCADA three-phase data time stamp to unify time sections.

DOI: 10.4236/epe.2020.124B008 76 Energy and Power Engineering

 S. T. Wu, Y. Li

based on the PMU at this moment. This article considers that the upload time of
SCADA data uploaded at this time is at 10 milliseconds.

3.3. Analysis of Data Upload Frequency Compatible Methods

On the issue of different upload frequencies. First, this article understands that
the upload frequency of the three-phase data of the SCADA system is on the
second level, which is assumed to be 2 seconds, and the upload frequency of the
three-phase data of the PMU system is on the millisecond level, which is also
assumed to be 10 ms here. In this way, we can know that in the same me-
dium-voltage power distribution system, the PMU has been uploaded dozens of
times by the SCADA system in the time interval for uploading data once. If they
cannot be fully utilized, this will waste a lot of high precision data. The PMU da-
ta is also a huge data loss for the state estimation, which makes the system dis-
patcher unable to grasp the system state information more accurately.
For this reason, this article also considers this problem under normal cir-
cumstances, that is, the change of power data in a short period of time is consi-
dered stable. For example, at the second or millisecond level, we can treat the
data change during this period as approximately linear, then we can use the li-
near change rule to handle the problem of inconsistent upload frequency.
Based on the above analysis, this paper intends to use Lagrangian low-order
linear interpolation algorithm to solve the problem of frequency compatibility of
three-phase data in different systems, and in the following description of this ar-
ticle, this method will be referred to as the interpolation algorithm for short. Be-
cause this method is divided into interpolation algorithm and extrapolation al-
gorithm, and the interpolation algorithm has a better effect on the internal in-
terpolation of the data, the extrapolation method is mainly aimed at the external
interpolation of the data. Therefore, this paper will use the interpolation algo-
rithm to enrich the historical database for frequency compatibility of historical
data; while the frequency compatibility of current or future time section data will
be extrapolated.
Regarding frequency compatible interpolation methods. Regarding frequency
compatible interpolation methods. As shown in Figure 2(a) below, as the inter-
polation interval decreases from T2 to T1, the interpolation result changes from
X2 to X1, and it can be seen that the accuracy also increases. So if you choose a
larger interpolation interval, the interpolation accuracy will be unsatisfactory,
and the minimum upload frequency of SCADA data is 2 seconds. Therefore, for
SCADA data frequency to match the data of PMU, the interpolation time interval
used by its interpolation algorithm can be 2 seconds. Then the calculation formula
for the phasor measurement of a specific SCADA node is shown in the following
Equation (1) and this paper considers that this size interval is also used for the
extrapolation algorithm, which is also feasible for improving the accuracy.
Tinsert − Tzone 2 p ,Tzone 2 Tinsert − Tzone1
= p ,Tinsert
M SCADA p ,Tzone1
M SCADA + M SCADA (1)
Tzone1 − Tzone 2 Tzone 2 − Tzone1

DOI: 10.4236/epe.2020.124B008 77 Energy and Power Engineering

S. T. Wu, Y. Li

Figure 2. Frequency compatible interpolation algorithm. (a) Example of interpolation

algorithm; (b) Examples of extrapolation algorithms.

In Equation (1): p is the phase sequence number, Tinsert is the target interpo-
lation time point, TZone1 is the first time point of the interpolation interval,
TZone 2 is the last time point of the interpolation interval, and TZone1 < Tinsert < TZone 2 .
About frequency compatible extrapolation algorithms. As shown in Figure
2(b), considering that the data on the current time section is not consistent with
the frequency, the SCADA system may not upload new data on this time section.
At this time, only the PMU data is not feasible for state estimation. Based on
this, in this scenario, this paper uses the extrapolation algorithm to quickly make
up for the SCADA measurement data that is not available in this time section,
thereby uniformly refreshing the frequency at this angle. At this time, the calcu-
lation formula for the interpolation result of the measured value of a certain
SCADA configuration point at a specified time is shown in the following Equa-
tion (2).
Tinsert − Tinsertzone 2 Tinsert − Tinsertzone1
= p ,Tinsert
M SCADA p ,Tinsertzone1
M SCADA + M SCADA
p ,Tinsertzone 2
(2)
Tinsertzone1 − Tinsertzone 2 Tinsertzone 2 − Tinsertzone1

In Equation (2): TinterZone1 is the first time point of the interpolation interval,
TinterZone 2 is the last time point of the interpolation interval, and
TinterZone1 < TinterZone 2 < Tinsert .

4. State Estimation Analysis

Because the three-phase asymmetry factor must be considered in the state esti-

DOI: 10.4236/epe.2020.124B008 78 Energy and Power Engineering

 S. T. Wu, Y. Li

mation of the distribution network, this article focuses on modeling the

three-phase line parameters of the distribution network before the state estima-
tion of the distribution network. The three-phase model of the specific line is
shown in Figure 3.
Among them, Zaa, Zbb, and Zcc are the self-impedance of the three-phase line;
Zab, Zbc, and Zca are the mutual impedance of the three-phase line; Ia, Ib, and Ic
are the currents flowing through the three-phase line.
The branch impedance matrix of a three-phase line is shown in Equation (3)
below:
 Z aa Z ab Z ca 
 
Z abc =  Z ab Zbb Zba  (3)
Z Zba Z cc 
 ca
The state estimation in this paper is to use the branch current method to es-
tablish the related estimation equation. That is to say, the state quantity selected
for the state estimation is the branch current amplitude and phase angle. At the
same time, considering the three-phase state, the state variable shape is shown in
the following Equation (4):
a
( I1,2 b
, I1,2 c
, I1,2 , β1,2
a
, β1,2
b
, β1,2
c
, , I ija , I ijb , I ijc , βija , βijb , βijc ,) (4)

In Equation (4): I ija is the branch a-phase current of the first node i, and the
end node is j; βija represents the phase angle of the corresponding branch cur-
rent. By the same token, the meanings of the other two related symbols can be
obtained.
In addition, because this article is for the state estimation of the distribution
network with SCADA system and PMU system, the data of the two different
systems are compatible, so pay attention to this compatibility process when per-
forming state estimation.
Then, in the state estimation of the distribution network, when the SCADA
system data and the PMU system data are simultaneously transmitted to the
state estimation data system, the state estimation process is directly performed.
Otherwise, the interpolation algorithm is used for making up the SCADA data,
and then the state estimation is performed.

Figure 3. Three-phase line model of distribution network.

DOI: 10.4236/epe.2020.124B008 79 Energy and Power Engineering

S. T. Wu, Y. Li

In summary, the measurement equation matrix for state estimation in this

paper is shown in Equation (5) below:
 Pija,( PMU , SCADA)   h1 ( I1,2a b
, I1,2 c
, I1,2 , β1,2
a
, β1,2
b
, β1,2
c
, , I ija , I ijb , I ijc , βija , βijb , βijc ,) 
 b   
 Pij ,( PMU , SCADA)   h1 ( I1,2 , I1,2 , I1,2 , β1,2 , β1,2 , β1,2 , , I ij , I ij , I ij , βij , βij , βij ,) 
a b c a b c a b c a b c

 Pc   h ( I a , I b , I c , β a , β b , β c , , I a , I b , I c , β a , β b , β c ,) 
 ij ,( PMU , SCADA)   1 1,2 1,2 1,2 1,2 1,2 1,2 ij ij ij ij ij ij

 Pi ,(a PMU , SCADA)   h1 ( I1,2a b
, I1,2 c
, I1,2 , β1,2
a
, β1,2
b
, β1,2
c
, , I ija , I ijb , I ijc , βija , βijb , βijc ,) 
 b   
 Pi ,( PMU , SCADA)   h1 ( I1,2 , I1,2 , I1,2 , β1,2 , β1,2 , β1,2 , , I ij , I ij , I ij , βij , βij , βij ,) 
a b c a b c a b c a b c

 Pc   h ( I a , I b , I c , β a , β b , β c , , I a , I b , I c , β a , β b , β c ,) 
 i ,( PMU , SCADA)   1 1,2 1,2 1,2 1, 2 1,2 1,2 ij ij ij ij ij ij

     
 Qa  =  h ( I a , I b , I c , β a , β b , β c , , I a , I b , I c , β a , β b , β c ,)  + v (5)
 ij ,( PMU , SCADA)   1 1,2 1,2 1,2 1,2 1,2 1,2 ij ij ij ij ij ij 
     
 a   
 Qi ,( PMU , SCADA)   h1 ( I1,2 , I1,2 , I1,2 , β1,2 , β1,2 , β1,2 , , I ij , I ij , I ij , βij , βij , βij ,) 
a b c a b c a b c a b c

     
 a   
 U i ,( PMU , SCADA)   h1 ( I1,2 , I1,2 , I1,2 , β1,2 , β1,2 , β1,2 , , I ij , I ij , I ij , βij , βij , βij ,) 
a b c a b c a b c a b c

     
 a   
 I ij ,( PMU , SCADA)   h1 ( I1,2
a b
, I1,2 c
, I1,2 , β1,2
a
, β1,2
b
, β1,2
c
, , I ija , I ijb , I ijc , βija , βijb , βijc ,) 
   
     
   
In Equation (5): Pija,( PMU , SCADA) represents PMU or SCADA active power
measurement in a-phase of branch, i and j represent the node numbers;
Pi ,a( PMU , SCADA) represents the a-phase node load measurement; I ija,( PMU , SCADA) is
the branch current measurement; U ia( PMU , SCADA) is the node voltage measure-
ment. Similarly, the meaning of other symbols can be obtained.
Finally, the constructed state estimation equations are solved using the
weighted least square method.

5. Example Analysis and Conclusion

The simulation of the example passed the IEEE33 node system platform, and
the method in this paper was verified by using the 50-second power flow cal-
culation method. It is assumed in the simulation that the upload frequency of
the PMU is 1 second (actually milliseconds), and the upload frequency of the
SCADA data is 10 seconds (actually 2 seconds). In order to understand the in-
terpolation effect in detail, this article sets a node as a SCADA measurement de-
vice, and assumes that the load changes and calculates the power flow. The ob-
tained data is used as the full SCADA measurement value (that is, the data under
the assumption that SCADA can be measured every second), and then every 10
seconds is used as a piecewise linear simulation interval, and interpolation is
used to solve the interpolation. Obtain the simulation results, that is, the effect
between the full measurement value and the interpolation simulation is shown
in Figure 4. Use extrapolation to solve the interpolation, the fusion effect is
shown in Figure 5.
Through analysis and calculation, the average error of the interpolated simu-
lation value is 0.71%, and the average error of the extrapolated simulation value

DOI: 10.4236/epe.2020.124B008 80 Energy and Power Engineering

 S. T. Wu, Y. Li

Figure 4. Interpolation simulation analysis.

Figure 5. Extrapolation simulation analysis.

Figure 6. State estimation with data compatibility.

is 1.92%. This paper considers that the error of both methods is acceptable, but
the interpolation method is better than the extrapolation method. Therefore, in
the calculation process, the solution obtained by the extrapolation method can
be corrected by interpolation.

DOI: 10.4236/epe.2020.124B008 81 Energy and Power Engineering

S. T. Wu, Y. Li

Use the above results for state estimation analysis. SCADA data is added with
a Gaussian error of 2%, and PMU data is added with a Gaussian error of 0.1%.
The PMU is configured at nodes 4, 6, 8, 12, 17, 19, 23, and the remaining nodes
are configured with SCADA systems. Compare the effects of full measurement
and interpolation, as shown in Figure 6. The results show that the state estima-
tion under interpolation is not much different from the state estimation results
under full measurement, which basically meets the requirements. The average
error between the two is about 0.1%. The error curves basically match. It shows
that the method in this paper can be applied to state estimation.

Acknowledgements
This work was financially supported by the National key research and development
plan: Intelligent Power Network Technology and Equipment (2017YFB0902800).

Conflicts of Interest
The authors declare no conflicts of interest regarding the publication of this pa-
per.

References
[1] Wang, K., Mu, G. and Chen, X. (2001) Precision Improvement and Pmu Placement
Studies on State Estimation of a Hybrid Measurement System with Pmus. Chinese
Society for Electrical Engineering, 21, 29-33.
[2] Van Cutsem, T. and Ribbens-Pavella, M. (1983) Critical Survey of Hierarchical Me-
thods for State Estimation of Electric Power Systems. IEEE Transactions on Power
Apparatus and Systems, 102, 3415-3424. https://doi.org/10.1109/TPAS.1983.317838
[3] Alves, G.O., et al. (2016) Distribution System State Estimation Using Phasor Mea-
surement Units. 10th International Conference on Compatibility, Power Electronics
and Power Engineering (CPE-POWERENG), Bydgoszcz, 29 June-1 July 2016,
2166-9546.
[4] Pau, M., Pegoraro, P.A. and Sulis, S. (2013) Efficient Branch-Current-Based Distri-
bution System State Estimation Including Synchronized Measurements. IEEE
Transactions on Instrumentation & Measurement, 62, 2419-2429.
https://doi.org/10.1109/TIM.2013.2272397
[5] Liu, J., Ponci, F., Monti, A., Muscas, C., Pegoraro, P.A. and Sulis, S. (2014) Optimal
Meter Placement for Robust Measurement Systems in Active Distribution Grids.
IEEE Transactions on Instrumentation and Measurement, 63, 1096-1105.
https://doi.org/10.1109/TIM.2013.2295657
[6] Alimardani, A., Therrien, F., Atanackovic, D., Jatskevich, J. and Vaahedi, E. (2015)
Distribution System State Estimation Based on Nonsynchronized Smart Meters.
IEEE Transactions on Smart Grid, 6, 2919-2928.
https://doi.org/10.1109/TSG.2015.2429640
[7] Gómez-Expósito, A., Gómez-Quiles, C. and Džafić. I. (2015) State Estimation in
Two Time Scales for Smart Distribution Systems. IEEE Transactions on Smart Grid,
6, 421-430. https://doi.org/10.1109/TSG.2014.2335611
[8] Ni, D., Zhang, W., Yu, B. and Gong, W. (2011) A New Algorithm for Power System
State Estimation with PMU Measurements. International Conference on Mecha-

DOI: 10.4236/epe.2020.124B008 82 Energy and Power Engineering

 S. T. Wu, Y. Li

tronic Science, Electric Engineering and Computer, Jilin, 19-22 August 2011,
114-117. https://doi.org/10.1109/MEC.2011.6025413
[9] Saadabadi, H. and Dehghani, M. (2016) Large-Scale Power Systems State Estimation
Using PMU and SCADA Data. 2016 24th Iranian Conference on Electrical Engi-
neering (ICEE), Shiraz, 10-12 May 2016, 906-911.
https://doi.org/10.1109/IranianCEE.2016.7585649
[10] Primadianto, A. and Lu, C. (2017) A Review on Distribution System State Estima-
tion. IEEE Transactions on Power Systems, 32, 3875-3883.
https://doi.org/10.1109/TPWRS.2016.2632156
[11] Li, B. (2013) Research on Dynamic State Estimation of Power System under Mixed
Measurement. Ph.D. Thesis, Shanghai Jiaotong University, Shanghai.

DOI: 10.4236/epe.2020.124B008 83 Energy and Power Engineering