Distribution System State Estimation

and Smart Meter Analysis

Instructor: Dr. Zhaoyu Wang

GRA: Dingwei Wang
Department of Electrical and Computer Engineering
Iowa State University
Contents
• Introduction to State Estimation (SE)
- The concept of SE
- Conventional SE method
• Introduction to Distribution System State Estimation (DSSE)
- Transmission and distribution grids
- The concept of DSSE
- Conventional DSSE method
• DSSE Research Topics
- The challenges of DSSE
- Observability problem
- Metering device placement
- Unbalanced problem
- Topology configuration
- Renewable integration
- Robust DSSE methods
• Smart meter data analysis
• Conclusion 2

ECE
 The Concept of SE
What is the state in the power system
• In general, power system has normal, emergency, and restorative states.
• To monitor system states, different measurements from all parts of the system
need to be utilized.
• State estimation is a data processing algorithm for converting redundant meter
readings into an estimate of the state of an electric power system.

Fig. 1 State Diagram for Power System Operation [1]

[1] Gomez-Exposito A, Abur A. Power system state estimation: theory and implementation[M]. CRC press, 2004. 3

ECE
The Concept of SE
SE is a widely-used tool in transmission systems. In the transmission
system SE, voltage magnitudes and phase angles are considered the
states of systems.

Fig. 2 On-line Static Security Assessment: Functional Diagram [1] 4

ECE
The Concept of SE
Why is it important to use SE in the power system-

Various constraints make it impossible to have a good picture of the

power system [2]:

• Because of economical constraints, measurement devices can not

be installed everywhere, so the data is incomplete.

• Because of the meter malfunction and the communication

problem, the measurements are subject to error or loss, so the data
is inaccurate, unreliable, and delayed.

[2] H. Wang and N. N. Schulz, “A revised branch current-based distribution system state estimation algorithm and meter placement impact,” IEEE
Trans. Power Syst., vol. 19, no. 1, pp. 207–213, Feb. 2004 5

ECE
The Concept of SE

Types of Measurement Errors

• Random errors – depend on the class of precision of the

measurements (phasor measurement unit, smart meter, etc.).

• Intermittent errors – large noise or temporary failures due to

communication or meter malfunction.

• Systematic errors – deterioration of measurements due to age,

temperature, weather, and other environmental effects [3].

[3] Zhong, Shan, and Ali Abur, “Combined state estimation and measurement calibration,” IEEE Trans. Power Syst., vol. 20, no. 1, pp. 458–465, Feb.,
2005
6

ECE
Conventional SE Method

Traditionally, bus voltage magnitudes and phase angles have been

used as state variables in transmission systems. The basic equation
of SE can be written by:

𝑧𝑧1 ℎ1 𝑥𝑥1 , … , 𝑥𝑥𝑛𝑛 𝑒𝑒1

𝑧𝑧 = ⋮ = ⋮ + ⋮ = ℎ 𝑥𝑥 + 𝑒𝑒
𝑧𝑧𝑚𝑚 ℎ𝑚𝑚 𝑥𝑥1 , … , 𝑥𝑥𝑛𝑛 𝑒𝑒𝑚𝑚

Where 𝑥𝑥𝑛𝑛 is the state variable of bus n, 𝑧𝑧𝑚𝑚 is the m-th real measurements,
ℎ𝑚𝑚 is the nonlinear measurement function to connect x and z, and e is the
measurement error.

ECE
Conventional SE Method

The problem can be formulated as a Weighted Least Square (WLS)

optimization method [1]:

� is the estimated state vector, 𝑇𝑇 is the matrix transposition

where 𝒙𝒙
operation, and W denotes the weight matrix that represents the
user’s confidence in the measured data. A widely-used choice for the
weight matrix is 𝑊𝑊 = 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝜎𝜎1−2 , … , 𝜎𝜎𝑚𝑚
−2
, where 𝜎𝜎𝑗𝑗−2 represents the
variance of the measurement error corresponding to the 𝑗𝑗𝑡𝑡𝑡 element
of measurement 𝑧𝑧.
8

ECE
Conventional SE Method

Conventionally, Gauss-Newton method has been applied to iteratively solve

the WLS problem [4]. The basic idea is to find a solution to the equation 𝛻𝛻𝐽𝐽 =
0:
𝜕𝜕𝜕𝜕
𝑯𝑯 𝒙𝒙 𝑘𝑘 =
𝜕𝜕𝒙𝒙(𝑘𝑘)
𝑇𝑇
𝑮𝑮 𝑘𝑘 = 𝑯𝑯 𝒙𝒙 𝑘𝑘 𝑾𝑾𝑾𝑾(𝒙𝒙 𝑘𝑘 )

−1 𝑯𝑯 𝑇𝑇
∆𝒙𝒙 𝑘𝑘 = 𝑮𝑮 𝑘𝑘 𝒙𝒙 𝑘𝑘 𝑾𝑾(𝒛𝒛 − 𝒉𝒉 𝒙𝒙 𝑘𝑘 )

𝒙𝒙 𝑘𝑘 + 1 = 𝒙𝒙 𝑘𝑘 + ∆𝒙𝒙 𝑘𝑘
Where 𝑯𝑯 is the Jacobian matrix with respect to the state variables and real measurements.
J denotes the objective function of the WLS problem. G(k) is the system gain matrix.

[4] F. F. Wu, “Power system state estimation: a survey,” International Journal of Electrical Power & Energy Systems, vol. 12, no. 2, pp. 80–87, Apr. 1990. 9

ECE
Conventional SE Method

𝑥𝑥 = [𝑉𝑉, 𝜃𝜃]
State Variables

Measurement Variables 𝑧𝑧 = [𝐼𝐼, 𝑉𝑉, 𝑃𝑃𝑏𝑏 , 𝑄𝑄𝑏𝑏 , 𝑃𝑃𝐿𝐿 , 𝑄𝑄𝐿𝐿 , 𝑃𝑃𝐿𝐿𝑠𝑠 , 𝑄𝑄𝐿𝐿𝑠𝑠 ]

𝜕𝜕𝑃𝑃𝑏𝑏 𝜕𝜕𝑃𝑃𝐿𝐿
…
Jacobian Matrix of the 𝐻𝐻 𝑥𝑥 = 𝜕𝜕𝑉𝑉 𝜕𝜕𝑉𝑉
State Equations 𝜕𝜕𝑄𝑄𝑏𝑏 𝜕𝜕𝑃𝑃𝐿𝐿
…
𝜕𝜕𝜃𝜃 𝜕𝜕𝜃𝜃

Weight matrix 1 For the pseudo load

𝑊𝑊𝑖𝑖𝑖𝑖 = �
10 For the actual measurements
𝐼𝐼: Line current measurements 𝑃𝑃𝑏𝑏 : Branch real power 𝑄𝑄𝐿𝐿 : Injection reactive power
𝑉𝑉: Voltage magnitudes 𝑄𝑄𝑏𝑏 : Branch reactive power 𝑃𝑃𝐿𝐿𝑠𝑠 : Pseudo injection real power
𝜃𝜃: Voltage angles 𝑃𝑃𝐿𝐿 : Injection real power 𝑄𝑄𝐿𝐿𝑠𝑠 : Pseudo injection reactive power

The voltage here can be used for both prediction features and verifications. 10

ECE
Contents
• Introduction to State Estimation (SE)
- The concept of SE
- Conventional SE method
• Introduction to Distribution System State Estimation (DSSE)
- Transmission and distribution grids
- The concept of DSSE
- Conventional DSSE method
• DSSE Research Topics
- The challenges of DSSE
- Observability problem
- Metering device placement
- Unbalanced problem
- Topology configuration
- Renewable integration
- Robust DSSE methods
• Smart meter data analysis
• Conclusion 11

ECE
Transmission Grid vs. Distribution Grid
Differences between transmission system and distribution system

Fig. 4 IEEE 34 Bus Test System.

Fig. 3 IEEE 24 Bus Test System.

Meshed topology Radial topology

Uni-directional power flows Bi-directional power flows
Balanced lines and loads Unbalanced lines and loads
Single phase analysis Three phase analysis 12

ECE
Transmission Grid vs. Distribution Grid
Differences between transmission system and distribution system
Items Transmission Network Distribution Network
The power supply situation is regional power
The general network topology is mesh-shaped and supply, the network topology in the region is
Topology
needs to be analyzed as a whole. radial, the closed-loop design between the
regions, and the open-loop operation.
The imbalance of the network is small and can The three-phase line parameters are
basically be ignored. It can be considered that the unbalanced, the R/X ratio fluctuates greatly,
Network imbalance three-phase line parameter balance and three- the three-phase load is unbalanced, and
phase load balance can be analyzed in single-phase there are single-phase and two-phase loads,
or positive sequence. which cannot be analyzed independently.
A small amount of real-time measurement, a
There are a large number of real-time large number of load pseudo-measurement,
SCADA measuring measurement devices and a small number of from the perspective of real-time
device pseudo-measurements, and the measurement measurement, the measurement redundancy
redundancy is high. is low, and the network value is generally
unobservable.
A typical network generally contains hundreds of A typical network generally contains 10,000
Network scale
buses to one or two thousand buses. to 100,000 nodes.
Mostly distributed DG distributed in the
Generally, thermal power, large-scale hydropower
Existing power feeder of the distribution network, the
and nuclear power generation, the output power is
plant output power fluctuates greatly, and has
basically stable and adjustable with the load.
certain controllability. 13

ECE
The Concept of DSSE
Why do we need to perform DSSE-

With complex interactions in distribution networks and rapid

growth of distributed energy resources (DER), electric vehicles,
SCADA, and advanced metering infrastructure (AMI), DSSE is
expected to become a significant function in monitoring and
power management of smart grids by estimating the high
accurate system states [5]-[6].

[5] “FERC staff report: Assessment of demand response and advanced metering - Dec. 2017.” [Online]. Available: https://www.ferc.gov/legal/
staff-reports/2017/DR-AM-Report2017.pdf.
[6] A. Primadianto and C. N. Lu, “A review on distribution system state estimation,” IEEE Trans. Power Syst., vol. 32, no. 5, pp. 3875–3883, Sep. 2017.

14

ECE
Transmission System SE vs Distribution System SE

• System Analysis Process:

• Transmission: Single phase analysis and one-line diagram of the system is
configured.

• Distribution: Unbalanced three-phase analysis and power flow constraints are

necessary.

• Data Availability:
• Transmission: Data is over-determined (the number of available measurements is
more than the number of estimations).

• Distribution: As the number of meter points is much lower in the distribution

network, most of the measurement data used in DSSE are pseudo measurements
data.

15

ECE
Transmission System SE vs Distribution System SE
• Measurement Jacobian Matrix:

Transmission: Distribution:

Reference: Sarada Devi, M. S. N. G., and G. Yesuratnam. "Comparison of State Estimation Process on Transmission and
Distribution Systems." Advances in Decision Sciences, Image Processing, Security and Computer Vision. Springer, Cham, 2020.
414-423. 16

ECE
 Distribution System Real-time Measurements

Fig. 5 Percent and numbers of Smart meter installations [7]

[7] Energy Information Administration. (2017) Annual Electric Power Industry Report. [Online]. Available:https://www.eia.gov/electricity/data/eia861/
17

ECE
The Process of DSSE
Like transmission system SE, DSSE is the process of inferring the
values of the distribution system’s state variables using a limited
number of measured data at certain locations in the system [8].

Fig. 6 DSSE function in smart grid environment [9].

[8] A. Monticelli, State estimation in electric power systems: a generalized approach. Springer Science & Business Media, 1999.
[9] K. Dehghanpour, Z. Wang, J. Wang, Y. Yuan and F. Bu, "A Survey on State Estimation Techniques and Challenges in Smart Distribution Systems," in IEEE
18
Transactions on Smart Grid, vol. 10, no. 2, pp. 2312-2322, March 2019.

ECE
The Method of DSSE

The selection of state variables in DSSE is separated into two

categories:

1. Voltage-Based DSSE [10]-[12]

2. Branch Current-Based SE (BCSE) [13]-[15]

[10] M. E. Baran and A. W. Kelley, “State estimation for real-time monitoring of distribution systems,” IEEE Trans. Power Syst., vol. 9, no. 3, pp. 1601–1609, Aug.
1994
[11] D. A. Haughton and G. T. Heydt, “A linear state estimation formulation for smart distribution systems,” IEEE Trans. Power Syst., vol. 28, no. 2, pp. 1187–
1195, May 2013.
[12] C. N. Lu, J. H. Tang, and W. H. E. Liu, “Distribution system state estimation,” IEEE Trans. Power Syst., vol. 10, no. 1, pp. 229–240, Feb. 1995.
[13] M. E. Baran and A. W. Kelley, “A branch-current-based state estimation method for distribution systems,” IEEE Trans. Power Syst., vol. 10, no. 1, pp. 483–
491, Feb. 1995.
[14] M. Pau, P. A. Pegoraro, and S. Sulis, “Efficient branch-current-based distribution system state estimation including synchronized measurements,” IEEE
Trans. Instrum. Meas., vol. 62, no. 9, pp. 2419–2429, Sep. 2013.
[15] M. E. Baran, J. Jung, and T. E. McDermott, “Including voltage measurements in branch current state estimation for distribution systems,” In IEEE Power &
Energy Society General Meeting, pp. 1–5, Jul. 2009. 19

ECE
The Method of DSSE
1. Voltage-Based DSSE [10]-[12]
• State variables: node voltage magnitude and angle.
• Equations: WLS approach with PF, similar to the conventional SE
method.
• Shortage: High computational complexity, mainly used for meshed
networks, sensitive to line parameters. May not work satisfactorily for
networks with a high R/X ratio [16].
2. Branch Current-Based SE (BCSE) [13]-[15]
• State variables: branch current and angle.
• Equations: WLS approach and can be expressed using the rectangular or
polar form.
• BCSE is more insensitive to line parameters than the conventional node-
voltage-based SE methods [9] and has better computation speed and
memory usage [2].
[16] Mohamed Ben Ahmed and Anouar Abdelhakim Boudhir. 2018. Innovations in Smart Cities and Applications: Proceedings of the 2nd Mediterranean
Symposium on Smart City Applications (1st ed.). Springer Publishing Company, Incorporated. 20

ECE
Branch Current-Based SE Algorithm
The three-phase branch current, also known as the state variables
of the system, x can be expressed as:
ph , real ph , real ph , real ph ,im ph ,im ph ,im T
x = [I 1 ...I l ...I N ,I
1 ...I l ...I N ]
𝑝𝑝𝑝,𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑝𝑝𝑝,𝑖𝑖𝑖𝑖
Where 𝐼𝐼𝑙𝑙 and 𝐼𝐼𝑙𝑙 represent the real and imaginary parts
of the three-phase branch current at branch 𝑙𝑙, respectively and 𝑁𝑁
represents the number of branches. The compact form can be
expanded as:
 I la ,real   I la ,im 
 b ,real   b ,im 
I lph ,real =  Il  I lph ,im =  Il 
 I c ,real   I c ,im 
 l   l  21

ECE
Branch Current-Based SE Algorithm
The system measurements considered are power, current
magnitude, and voltage magnitude measurements. They are derived
as follows:

(1) Power flow Measurements:

*
 Pl + jQl 
ph
ph ph
ph ,real ph ,im
Il 
= ph ,k
= I l + jI l
 Vi 
𝑝𝑝𝑝,𝑘𝑘
where 𝑉𝑉𝑖𝑖 is the estimated node voltage at the k-th iteration.
The reason for converting all power measurements is to have
linear relationships between the equivalent currents and the
state variables as it can be observed from this equation.
22

ECE
Branch Current-Based SE Algorithm

(2) Current Magnitude Measurements:

ph ,real 2 ph ,im 2
=Il ph
( Il ) + ( Il )

(3) Voltage Magnitude Measurements:

N
V=
j Vs − ∑ Z l I l
ph ph ph ph

l =1
𝑝𝑝𝑝
where 𝑉𝑉𝑗𝑗 is the substation voltage.

23

ECE
Branch Current-Based SE Algorithm
Detailed Algorithm:

-Step 1 Initialization:
 Set the initial value of voltage at every node, such as 1 pu.
 Backward Step: Using the injected power at every node,
the values of state variables (branch current magnitudes and
phase angles) are computed starting from the end of
networks.

-Step 2 WLS
 Using the WLS method, the state variable increments are
obtained.
 Update the value of state variables.

24

ECE
Branch Current-Based SE Algorithm
Detailed Algorithm:

-Step 3 Forward Step:

 Using the new values of state variables, the values of nodal
voltages are calculated starting from the substation.

-Step 4 Convergence Analysis

 If the increments are smaller than the tolerance: stop. If not,
return to step 2.

25

ECE
Block Diagram for solving WLS
problem:

• Start by sitting the iteration k = 0.

• Initialize the node voltages.

• Find the gain matrix G(x).

• After obtaining the branch
current estimate, update the node
voltages using the forward
sweeping approach.
• Check for convergence, If no,
update k = k + 1. Else, stop.
26

ECE
Contents
• Introduction to State Estimation (SE)
- The concept of SE
- Conventional SE method
• Introduction to Distribution System State Estimation (DSSE)
- Transmission and distribution grids
- The concept of DSSE
- Conventional DSSE method
• DSSE Research Topics
- The challenges of DSSE
- Observability problem
- Metering device placement
- Unbalanced problem
- Topology configuration
- Renewable integration
- Robust DSSE methods
• Smart meter data analysis
• Conclusion 27

ECE
The Challenges of DSSE
Compared to transmission system SE, DSSE is facing some unique
challenges due to the characteristics of the system[2]. These
challenges include:

• Low observability due to the lack of measurement device placements

• Higher R/X ratio

• Three-phase unbalanced system

• Communication issues and network topology identification problem

• Renewable energy and EV integration

• Cyber-security issues
28

ECE
Observability Problem
• Unlike transmission systems with a high data redundancy level, the
distribution systems are generally undetermined with low observability.

• “Observability” refers to the system operator’s ability to solve the SE

problem. That depends on the number and location of metering devices.

• Observability problem is one of the main challenges in applying

transmission SE techniques to distribution systems directly [2].

• In the traditional WLS-based SE method, the number of measurements

must be larger than the estimated states.

• The bad/missing measurement data also causes the observability problem.

29

ECE
Observability Problem
Distribution systems can be divided into three groups according to
observability: fully observable systems, partially observable
systems, and fully unobservable systems.

Fig. 9 Distribution Systems with different observability.

30

ECE
Observability Problem
Distribution System Real-time Measurements

Fig. 10 Percent of Residential Smart meter installation rate by state, 2016 [7].
31

ECE
Observability Problem

How to solve the observability problem-

• Observability problem is addressed by generating pseudo-

measurements when real measurements are unavailable [17].

• Pseudo-measurements are artificially-generated data points based

on the data history of the distribution systems.

• Use of poor accurate pseudo-measurements will introduce high

variance levels in the weight matrix, which could even lead to ill-
conditioning of the DSSE problem [17].

[17] A. Angioni, T. Schlosser, F. Ponci, and A. Monti, “Impact of pseudo-measurements from new power profiles on state estimation in low-voltage grids,” IEEE Trans. Instrum.
Meas., vol. 65, no. 1, pp. 70–77, Jan. 2016.
32

ECE
Observability Problem

The existing data-driven pseudo-measurement methods can be

roughly separated into two categories,

• Probabilistic and Statistical Approaches:

These methods employ spatial/temporal correlation and historic probability
distribution data to generate reasonable pseudo-measurements and assess their
uncertainty [18]-[21].

• Encouraged to read:
[18] C. Muscas, M. Pau, P. A. Pegoraro, and S. Sulis, “Effects of measurements and pseudo-measurements correlation in distribution system state estimation,”
IEEE Trans. Instrum. Meas., vol. 63, no. 12, pp. 2813–2823, Dec. 2014.
[19] A. K. Ghosh, D. L. Lubkeman, M. J. Downey, and R. H. Jones, “Distribution circuit state estimation using a probabilistic approach,” IEEE Trans. Power Syst., vol.
12, no. 1, pp. 45–51, Feb. 1997.
[20] R. Singh, B. C. Pal, and R. A. Jabr, “Statistical representation of distribution system loads using Gaussian mixture model,” IEEE Trans. Power Syst., vol. 25, no.
1, pp. 29–37, Feb. 2010.
[21] R. Singh, B. C. Pal, and R. A. Jabr, “Distribution system state estimation through Gaussian mixture model of the load as pseudo-measurement,” IET Gener.
Transm. Distrib., vol. 4, no. 1, pp. 50–59, Jan. 2009.
33

ECE
Observability Problem

Existing data-driven pseudo-measurement method can be roughly

separated into two categories:

• Learning-Based Approaches:
Multiple machine learning algorithms have also been utilized to generate
active/reactive power pseudo-measurement and uncertainty assessment [22]-
[26].

• Encouraged to read:
[22] B. P. Hayes, J. K. Gruber, and M. Prodanovic, “A closed-loop state estimation tool for MV network monitoring and operation,” IEEE Trans. Smart Grid, vol. 6,
no. 4, pp. 2116–2125, Jul. 2015.
[23] D. Gerbec, S. Gasperic, I. Smon, and F. Gubina, “Allocation of the load profiles to consumers using probabilistic neural networks,” IEEE Trans. Power Syst.,
vol. 20, no. 2, pp. 548–555, May 2005.
[24] E. Manitsas, R. Singh, B. C. Pal, and G. Strbac, “Distribution system state estimation using an artificial neural network approach for pseudo measurement
modeling,” IEEE Trans. Power Syst., vol. 27, no. 4, pp. 1888–1896, Nov. 2012.
[25] Y. Yuan, K. Dehghanpour, F. Bu, and Z. Wang, "A Multi-Timescale Data-Driven Approach to Enhance Distribution System Observability," IEEE Transactions on
Power Systems, vol. 34, no. 4, pp. 3168-3177, July 2019.
[26] K. Dehghanpour, Y. Yuan, Z. Wang and F. Bu, "A Game-Theoretic Data-Driven Approach for Pseudo-Measurement Generation in Distribution System
State Estimation," in IEEE Transactions on Smart Grid.
34

ECE
Metering Device Placement
Optimizing the location of meters in distribution systems is a significant
subject for research, given the size of the system and potentially limited
financial resources [9].

Objective Function Constraints Solution Algorithm

Meter cost [27] Estimation accuracy Heuristic search
Estimation accuracy [28] Meter number Mixed integer semidefinite
optimization
Network observability [29] NA Heuristic search
Meter cost & estimation Estimation accuracy Multi-Objective evolutionary
accuracy [30]

[27] M. E. Baran, J. Zhu, and A. W. Kelley, “Meter placement for real-time monitoring of distribution feeders,” IEEE Trans. Power Syst., vol. 11, no. 1, pp. 332–337, Feb. 1996.
[28] T. C. Xygkis, G. N. Korres, and N. M. Manousakis, “Fisher information based meter placement in distribution grids via the d-optimal experimental design,” IEEE Trans. Smart Grid, vol. 9,
no. 2, pp. 1452–1461, Mar. 2018.
[29] B. Brinkmann and M. Negnevitsky, “A probabilistic approach to observability of distribution networks,” IEEE Trans. Power Syst., vol. 32, no. 2, pp. 1169–1178, Mar. 2017.
[30] S. Prasad and D. M. V. Kumar, “Trade-offs in PMU and IED deployment for active distribution state estimation using multi-objective evolutionary algorithm,” IEEE Trans. Instrum. Meas.,
vol. 67, no. 6, pp. 1298–1307, Jun 2018.

35

ECE
Three Phase unbalanced Problem
In distribution systems, loads can be three-phase, two-phase, or single-phase.
Hence it is desirable to use a three-phase model in DSSE [14]. The basic WLS SE
method was adapted for three-phase analysis to address the phase unbalanced
problem [31].

Fig. 11 Three-phase branch model [30].

[31] U. Kuhar, M. Pantos, G. Kosec, and A. Svigelj, “The impact of model and measurement uncertainties on a state estimation in three-phase
distribution networks,” to appear in IEEE Trans. Smart Grid.
36

ECE
Three Phase unbalanced Problem - Solution
• Functions that relate measurements to the vector of state variables are
developed from a three-phase branch model.

• The BCSE method that was demonstrated earlier is modeled in three phases.

• However, according to [30], the measurement uncertainties will impact

estimation accuracy for different estimators, such as LAV, WLS, LMS, and
SHGM.

• Examples using different system models to achieve three-phase distribution

state estimation can be found in [32-34]

[32] Langner, Andre L., and Ali Abur. "Formulation of three-phase state estimation problem using a virtual reference." IEEE
Transactions on Power Systems 36.1 (2020): 214-223.
[33] A. Majumdar and B. C. Pal, "A three-phase state estimation in unbalanced distribution networks with switch modelling," 2016 IEEE
First International Conference on Control, Measurement and Instrumentation (CMI), 2016, pp. 474-478, doi:
10.1109/CMI.2016.7413793.
[34] F. Magnago, L. Zhang and R. Nagarkar, "Three phase distribution state estimation utilizing common information model," 2015
IEEE Eindhoven PowerTech, 2015, pp. 1-6, doi: 10.1109/PTC.2015.7232515.
37

ECE
High R/X Ratio Problem

• Distribution line with a high R/X ratio is another challenge for

SE [6].
𝜕𝜕𝑃𝑃𝑏𝑏 𝜕𝜕𝑃𝑃𝐿𝐿
…
• Recall the Jacobian Matrix H: 𝐻𝐻 𝑥𝑥 = 𝜕𝜕𝑉𝑉 𝜕𝜕𝑉𝑉
𝜕𝜕𝑄𝑄𝑏𝑏 𝜕𝜕𝑃𝑃𝐿𝐿
…
𝜕𝜕𝜃𝜃 𝜕𝜕𝜃𝜃

• The high R/X ratio results in the ill‐conditioning of H matrix. In

the transmission system, the off‐diagonal elements ∂P/∂V and
∂Q/∂θ are neglected, that is ∂P/∂V ≈ ∂Q/ ∂θ ≈ 0, because of weak
coupling. With a high R/X ratio and strong coupling, the
off‐diagonal elements cannot be discarded in the distribution
system.
38

ECE
High R/X Ratio Problem -Solution

• To avoid the ill‐conditioning of H matrix, the P and Q measurements

are handled by converting into equivalent current measurement I and
making the H matrix independent of state variables equation.

• The equivalent current measurement can be expressed with voltage or a

current variable for every iteration k.
∗
𝑘𝑘 𝑃𝑃 + 𝑗𝑗𝑗𝑗
𝐼𝐼𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑎𝑎,𝑏𝑏,𝑐𝑐 =
𝑉𝑉𝑘𝑘

• In this case, to address this challenge, branch current (BCSE) have

been adopted as state variables, which turns out to be a more natural
way of DSSE formulation [9].
[35] Bhattar, Poornachandratejasvi Laxman, Naran M. Pindoriya, and Anurag Sharma. "A combined survey on distribution
system state estimation and false data injection in cyber‐physical power distribution networks." IET Cyber‐Physical Systems:
Theory & Applications 6.2 (2021): 41-62. 39

ECE
Topology Configuration
DSSE relies on the basic assumption that we know the exact network model so
that we can write the measurement functions h(x). Hence, it is necessary to
perform topology configuration process to identify the current topology.

Fig. 30 DSSE and possible associated applications into a DMS.

40

ECE
Topology Configuration
The existing topology configuration method can be roughly separated into two
categories:

• System Identification Approaches:

These methods assume the basic topology of the network is known to the system
operator. However, due to local events, such as faults, line disconnections,
switching events, etc, the basic topology will undergo local changes over time.
[36]-[40].

[36] G. N. Korres and N. M. Manousakis, “A state estimation algorithm for monitoring topology changes in distribution systems,” in Proc. IEEE
Power Energy Soc. Gen. Meeting, San Diego, CA, USA, Jul. 2012, pp. 1–8.
[37] M. E. Baran, J. Jung, and T. E. McDermott, “Topology error identification using branch current state estimation for distribution systems,” In
IEEE Transmission & Distribution Conference & Exposition: Asia and Pacific, pp. 1–4, Oct. 2009.
[38] D. Singh, J. P. Pandey, and D. S. Chauhan, “Topology identification, bad data processing, and state estimation using fuzzy pattern matching,”
IEEE Trans. Power Syst., vol. 20, no. 3, pp. 1570–1579, Aug. 2005.
[39] G. Cavraro and R. Arghandeh, “Power distribution network topology detection with time-series signature verification method,” IEEE Trans.
Power Syst., vol. 33, no. 4, pp. 3500–3509, Jul. 2018.
[40] W. Luan, J. Peng, M. Maras, J. Lo, and B. Harapnuk, “Smart meter data analytics for distribution network connectivity verification,” IEEE Trans.
Smart Grid, vol. 6, no. 4, pp. 1964–1971, Jul. 2015.
41

ECE
Topology Configuration
The existing topology configuration method can be roughly separated into two
categories:

• Topology learning Approaches:

These methods assume that the system operator has very limited or no
knowledge of the basic topology of the network. Hence, the objective is to learn
the network’s topology using nodal and branch measurements [41]-[44].

[41] M. Babakmehr, M. G. Simões, M. B. Wakin, and F. Harirchi, “Compressive sensing-based topology identification for smart grids,” IEEE Trans.
Ind. Informat., vol. 12, no. 2, pp. 532–543, Apr. 2016.
[42] Y. Weng, Y. Liao, and R. Rajagopal, “Distributed energy resources topology identification via graphical modeling,” IEEE Trans. Power Syst., vol.
32, no. 4, pp. 2682–2694, Jul. 2017.
[43] S. J. Pappu, N. Bhatt, R. Pasumarthy, and A. Rajeswaran, “Identifying topology of low voltage distribution networks based on smart meter
data,” IEEE Trans. Smart Grid, vol. 9, no. 5, pp. 5113–5122, Sep. 2018.
[44] J. Yu, Y. Weng, and R. Rajagopal, “PaToPa: A data-driven parameter and topology joint estimation framework in distribution grids,” IEEE Trans.
Power Syst., vol. 33, no. 4, pp. 4335–4347, Jul. 2018.

42

ECE
Renewable Energy Integration

• The higher penetration of renewable power resources

introduces a higher level of uncertainty in DSSE.

• With the integration of DER and EV charging station, typical

load patterns are becoming unreliable and uncertain.

• The non-Gaussian distribution of renewable generation would

adversely affect WLS-based DSSE methods [9].

• Fast changes in system states can result in unreasonable errors

of the WLS-based DSSE [45].

[45] Y. Weng, R. Negi, C. Faloutsos, and M. D. Ilic, “Robust data-driven state estimation for smart grid,” IEEE Trans. Smart Grid, vol. 8, no. 4,
pp. 1956–1967, Jul. 2017.
43

ECE
Renewable Energy Integration

• Probabilistic methods represent the major group of techniques

for modeling the impacts of renewable uncertainty on DSSE
[9].

• Use GMM technique to obtain the non-Gaussian

distribution of renewable power [46].

• Use Beta distribution function to generate renewable

pseudo-measurement [47].

[46] G. Valverde, A. T. Saric, and V. Terzija, “Stochastic monitoring of distribution networks including correlated input variables,” IEEE Trans. Power
Syst., vol. 28, pp. 246–255, Feb. 2013.
[47] A. Angioni, T. Schlosser, F. Ponci, and A. Monti, “Impact of pseudo measurements from new power profiles on state estimation in low voltage
grids,” IEEE Trans. Instrum. Meas., vol. 65, no. 1, pp. 70–77, Jan. 2016.

44

ECE
Cyber Security
Due to the vulnerability of the power system against cyber-
attacks has been observed in practice, several common types of
cyber-attack related to SE have been modeled in the literature:

• False data injection [48]-[51]

• Topology attacks [52]-[53]
• Data privacy attacks [54]
[48] Q. Yang et al., “On false data-injection attacks against power system state estimation: Modeling and countermeasures,” IEEE Trans. Parallel Distrib.
Syst, vol. 25, no. 3, pp. 717–729, Mar. 2014.
[49] S. Li, Y. Yilmaz, and X. Wang, “Quickest detection of false data injection attack in wide-area smart grids,” IEEE Trans. Smart Grid, vol. 6, no. 6, pp.
2725–2735, Nov. 2015.
[50] J. Liang, L. Sankar, and O. Kosut, “Vulnerability analysis and consequences of false data injection attack on power system state estimation,” IEEE Trans.
Power Syst., vol. 31, no. 5, pp. 3864–3872, Sep. 2016.
[51] Y. Chakhchoukh and H. Ishii, “Enhancing robustness to cyber-attacks in power systems through multiple least trimmed squares state estimations,”
IEEE Trans. Power Syst., vol. 31, no. 6, pp. 4395–4405, Nov. 2016.
[52] Y. Chakhchoukh and H. Ishii, “Coordinated cyber-attacks on the measurement function in hybrid state estimation,” IEEE Trans. Power Syst., vol. 30,
no. 5, pp. 2487–2497, Sep. 2015.
[53] J. Zhang and L. Sankar, “Physical system consequences of unobservable state-and-topology cyber-physical attacks,” IEEE Trans. Smart Grid, vol. 7, no.
4, pp. 2016–2025, Jul. 2016.
[54] H. Li, L. Lai, and W. Zhang, “Communication requirement for reliable and secure state estimation and control in smart grid,” IEEE Trans. Smart Grid,
vol. 2, no. 3, pp. 476–486, Sep. 2011. 45

ECE
Robust DSSE Methods

• WLS estimator is a quadratic form of the maximum likelihood

estimator and can be stated as the minimization of the weighted
sum of squares. It is a fast and widely-used mathematical
formulation. However, WLS is sensitive to bad data.

• To handle the uncertainty of measurement data, alternative

mathematical formulations have been proposed to increase the
robustness of the state estimator.

46

ECE
Robust DSSE Methods
• Least Absolute Value estimator (LAV): LAV estimator is based
on the minimization of the 𝐿𝐿1 norm of weighted measurement
residual [55].

• Schweppe Huber Generalised M (SHGM): SHGM estimator is

an estimator that combines both WLS and WLAV [56].

• Matrix completion: MC Utilizes a direct connection between

the row and column elements within the matrix. The method
can estimate missing values in the smart meter measurement
matrices caused by low observability [57-58].
[55] R. Jabr, B. Pal, and R. Singh, “Choice of estimator for distribution system state estimation,” IET Generation, Transmission & Distribution, vol. 3, no. 7, pp.
666–678, Jul. 2009.
[56] Donti, Priya L., et al. "Matrix completion for low-observability voltage estimation." IEEE Transactions on Smart Grid 11.3, 2019, pp. 2520-2530.
[57] Zhang, Yingchen, et al. “State estimation in low-observable distribution systems using matrix completion.” National Renewable Energy Lab. (NREL), 2019.
[58] Liu, Bo, et al. "Robust matrix completion state estimation in distribution systems." 2019 IEEE Power & Energy Society General Meeting (PESGM). IEEE,
2019. 47

ECE
Smart Meter Data Analytics
Why is it important to perform SM data analytics-

• The widespread popularity of SMs enables an immense amount

of fine-grained electricity consumption data to be collected [59].

• High-resolution data from SM provide rich information to

understand the consumption behaviors of the consumers.

• SM data provides a unique opportunity to develop a data-

enabled modernized power system [60].

[59] Y. Wang, Q. Chen, T. Hong and C. Kang, "Review of Smart Meter Data Analytics: Applications, Methodologies, and Challenges," in
IEEE Transactions on Smart Grid, vol. 10, no. 3, pp. 3125-3148, May 2019.
[60] H. Sun, Z. Wang, J. Wang, Z. Huang, N. Carrington and J. Liao, "Data-Driven Power Outage Detection by Social Sensors," in IEEE
Transactions on Smart Grid, vol. 7, no. 5, pp. 2516-2524, Sept. 2016.
48

ECE
Smart Meter Data Analytics

Fig. 13 Number of publication about SM data Fig. 14 Number of publications in most popular
analytics。 journals.

• In 2010, the number of SM data analytics publications was at a low level.

• The number of publications increased rapidly from 2012.

49

ECE
Smart Meter Data Analytics

Fig. 15 Taxonomy of SM data analytics。

50

ECE
Conclusions

• The goal of DSSE is to infer the values of the distribution system’s state variables
using a limited number of measured data. The DSSE is a numerical process to
map data measurements to system state variables.

• Technically, conventional transmission level SE approaches cannot be directly

applied to the DSSE due to various challenges.

• To address these challenges, different methods are proposed for DSSE. Most
recent works are concentrated on using SM data analytics-based approaches to
improve the conventional DSSE.

• How to take advantage of massive SM data to enhance the efficiency of the

demand side has become an important topic.

51

ECE
Appendix I: Matlab Example of
Solving WLS Problem
A=importdata('H_data.m'); % import bus and line data with 5 columns, this is saved in another file
fbus=A(:,1); %From bus
tbus=A(:,2); %To bus
x=A(:,3); % Reactance in pu
Zm=A(:,4);% Impedance in pu
sig=A(:,5);% Susceptance in pu
buses=input('Enter the number of buses : \n');
B=input('Enter the reference bus : ');
for i=1:buses
H(i,fbus(i))=1/x(i);
H(i,tbus(i))=-1/x(i);
end
H(:,B)=[]; % Compute H matrix
for i=1:buses
R(i,i)=sig(i)^2; % Compute R Matrix
end
[nm,ns]=size(H);
if ns<nm
xest=[H'*R^-1*H]^-1*H'*R^-1*Zm; %Compute Gain matrix and forward sweep
else if ns==nm
xest=H^-1*Zm;
else if ns>nm
H'*[H*H']^-1*Zm;
end
end
end
Zt=H*xest; %True Value
e=Zm-Zt; %estimation error
fprintf(' The H Matrix \n');
disp(H);
fprintf(' The R inverse Matrix \n');
disp(R^-1);
fprintf('Estimated Value \n');
disp(xest);
fprintf('\n Measured value \n');
disp(Zm);
fprintf('\n True Value \n');
disp(Zt);
fprintf('\nError in equipment \n');
disp(e);
Display Results:
For 3 bus test case, use bus 1 as reference bus:

The H Matrix Measured value

-5.0000 0 0.6000
0 -2.5000 0.4000
-4.0000 4.0000 0.4050

The R inverse Matrix

10000 0 0 True Value
0 10000 0 0.7076
0 0 10000 0.1848
0.2705

Error in equipment
-0.1076
0.2152
0.1345
Appendix II: VBSE Matlab Code
Using WLS Method
• The IEEE 14 bus system is used in this example.

Bus data:

% |Bus | Type | Vsp | theta | PGi | QGi | PLi | QLi | Qmin | Qmax |
busdata14 = [ 1 1 1.060 0 0 0 0 0 0 0;
2 2 1.045 0 40 42.4 21.7 12.7 -40 50;
3 2 1.010 0 0 23.4 94.2 19.0 0 40;
4 3 1.0 0 0 0 47.8 -3.9 0 0;
5 3 1.0 0 0 0 7.6 1.6 0 0;
6 2 1.070 0 0 12.2 11.2 7.5 -6 24;
7 3 1.0 0 0 0 0.0 0.0 0 0;
8 2 1.090 0 0 17.4 0.0 0.0 -6 24;
9 3 1.0 0 0 0 29.5 16.6 0 0;
10 3 1.0 0 0 0 9.0 5.8 0 0;
11 3 1.0 0 0 0 3.5 1.8 0 0;
12 3 1.0 0 0 0 6.1 1.6 0 0;
13 3 1.0 0 0 0 13.5 5.8 0 0;
14 3 1.0 0 0 0 14.9 5.0 0 0;];
Line Data:

% | From | To | R | X | B/2 | X'mer |

% | Bus | Bus | pu | pu | pu | TAP (a) |
linedata14 =[1 2 0.01938 0.05917 0.0264 1
1 5 0.05403 0.22304 0.0246 1
2 3 0.04699 0.19797 0.0219 1
2 4 0.05811 0.17632 0.0170 1
2 5 0.05695 0.17388 0.0173 1
3 4 0.06701 0.17103 0.0064 1
4 5 0.01335 0.04211 0.0 1
4 7 0.0 0.20912 0.0 0.978
4 9 0.0 0.55618 0.0 0.969
5 6 0.0 0.25202 0.0 0.932
6 11 0.09498 0.19890 0.0 1
6 12 0.12291 0.25581 0.0 1
6 13 0.06615 0.13027 0.0 1
7 8 0.0 0.17615 0.0 1
7 9 0.0 0.11001 0.0 1
9 10 0.03181 0.08450 0.0 1
9 14 0.12711 0.27038 0.0 1
10 11 0.08205 0.19207 0.0 1
12 13 0.22092 0.19988 0.0 1
13 14 0.17093 0.34802 0.0 1 ];
Measurement Data:
% Vi - 1, Pi - 2, Qi - 3, Pij - 4, Qij - 5, Iij - 6; %------ Real Power Flow ------------- %
18 4 1.5708 1 2 64e-6;
% |Msnt |Type | Value | From | To | Rii | 19 4 0.7340 2 3 64e-6;
zdata14 = [ %---- Voltage Magnitude ------------% 20 4 -0.5427 4 2 64e-6;
1 1 1.06 1 0 9e-4; 21 4 0.2707 4 7 64e-6;
%-----------------------------------% 22 4 0.1546 4 9 64e-6;
%---- Real Power Injection ---------% 23 4 -0.4081 5 2 64e-6;
2 2 0.1830 2 0 1e-4; 24 4 0.6006 5 4 64e-6;
3 2 -0.9420 3 0 1e-4; 25 4 0.4589 5 6 64e-6;
4 2 0.00 7 0 1e-4; 26 4 0.1834 6 13 64e-6;
5 2 0.00 8 0 1e-4; 27 4 0.2707 7 9 64e-6;
6 2 -0.0900 10 0 1e-4; 28 4 -0.0816 11 6 64e-6;
7 2 -0.0350 11 0 1e-4; 29 4 0.0188 12 13 64e-6;
8 2 -0.0610 12 0 1e-4; %------------------------------------%
9 2 -0.1490 14 0 1e-4; %------ Real Power Flow ------------- %
%------------------------------------% 30 5 -0.1748 1 2 64e-6;
%---- Reactive Power Injection ------% 31 5 0.0594 2 3 64e-6;
10 3 0.3523 2 0 1e-4; 32 5 0.0213 4 2 64e-6;
11 3 0.0876 3 0 1e-4; 33 5 -0.1540 4 7 64e-6;
12 3 0.00 7 0 1e-4; 34 5 -0.0264 4 9 64e-6;
13 3 0.2103 8 0 1e-4; 35 5 -0.0193 5 2 64e-6;
14 3 -0.0580 10 0 1e-4; 36 5 -0.1006 5 4 64e-6;
15 3 -0.0180 11 0 1e-4; 37 5 -0.2084 5 6 64e-6;
16 3 -0.0160 12 0 1e-4; 38 5 0.0998 6 13 64e-6;
17 3 -0.0500 14 0 1e-4; 39 5 0.1480 7 9 64e-6;
%------------------------------------% 40 5 -0.0864 11 6 64e-6;
41 5 0.0141 12 13 64e-6;];
%--------------------------------------%
State Estimation using Weighted Least Square Method:
Parameter Settings:
num = 14; % IEEE - 14 bus system
ybus = ybusppg(num); % Get YBus
zdata = zdatas(num); % Get Measurement data
bpq = bbusppg(num); % Get B data
nbus = max(max(zdata(:,4)),max(zdata(:,5))); % Get number of buses
type = zdata(:,2); % Type of measurement, Vi - 1, Pi - 2, Qi - 3, Pij - 4, Qij - 5, Iij - 6
z = zdata(:,3); % Measuement values
fbus = zdata(:,4); % From bus
tbus = zdata(:,5); % To bus
Ri = diag(zdata(:,6)); % Measurement Error
V = ones(nbus,1); % Initialize the bus voltages
del = zeros(nbus,1); % Initialize the bus angles
E = [del(2:end); V]; % State Vector
G = real(ybus);
B = imag(ybus);

vi = find(type == 1); % Index of voltage magnitude measurements

ppi = find(type == 2); % Index of real power injection measurements
qi = find(type == 3); % Index of reactive power injection measurements
pf = find(type == 4); % Index of real powerflow measurements
qf = find(type == 5); % Index of reactive powerflow measurements

nvi = length(vi); % Number of Voltage measurements

npi = length(ppi); % Number of Real Power Injection measurements
nqi = length(qi); % Number of Reactive Power Injection measurements
npf = length(pf); % Number of Real Power Flow measurements
nqf = length(qf); % Number of Reactive Power Flow measurements

iter = 1;
tol = 5;
while(tol > 1e-4)

%Measurement Function, h
h1 = V(fbus(vi),1);
h2 = zeros(npi,1);
h3 = zeros(nqi,1);
h4 = zeros(npf,1);
h5 = zeros(nqf,1);

for i = 1:npi
m = fbus(ppi(i));
for k = 1:nbus
h2(i) = h2(i) + V(m)*V(k)*(G(m,k)*cos(del(m)-del(k)) + B(m,k)*sin(del(m)-del(k)));
end
end

for i = 1:nqi
m = fbus(qi(i));
for k = 1:nbus
h3(i) = h3(i) + V(m)*V(k)*(G(m,k)*sin(del(m)-del(k)) - B(m,k)*cos(del(m)-del(k)));
end
end

for i = 1:npf
m = fbus(pf(i));
n = tbus(pf(i));
h4(i) = -V(m)^2*G(m,n) - V(m)*V(n)*(-G(m,n)*cos(del(m)-del(n)) - B(m,n)*sin(del(m)-del(n)));
end

for i = 1:nqf
m = fbus(qf(i));
n = tbus(qf(i));
h5(i) = -V(m)^2*(-B(m,n)+bpq(m,n)) - V(m)*V(n)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n)));
end

h = [h1; h2; h3; h4; h5];

% Residue
r = z - h;
% Jacobian
% H11 - Derivative of V with respect to angles, All Zeros
H11 = zeros(nvi,nbus-1);

% H12 - Derivative of V with respect to V

H12 = zeros(nvi,nbus);
for k = 1:nvi
for n = 1:nbus
if n == k
H12(k,n) = 1;
end
end
end

% H21 - Derivative of Real Power Injections with Angles

H21 = zeros(npi,nbus-1);
for i = 1:npi
m = fbus(ppi(i));
for k = 1:(nbus-1)
if k+1 == m
for n = 1:nbus
H21(i,k) = H21(i,k) + V(m)* V(n)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n)));
end
H21(i,k) = H21(i,k) - V(m)^2*B(m,m);
else
H21(i,k) = V(m)* V(k+1)*(G(m,k+1)*sin(del(m)-del(k+1)) - B(m,k+1)*cos(del(m)-del(k+1)));
end
end
end
% H22 - Derivative of Real Power Injections with V
H22 = zeros(npi,nbus);
for i = 1:npi
m = fbus(ppi(i));
for k = 1:(nbus)
if k == m
for n = 1:nbus
H22(i,k) = H22(i,k) + V(n)*(G(m,n)*cos(del(m)-del(n)) + B(m,n)*sin(del(m)-del(n)));
end
H22(i,k) = H22(i,k) + V(m)*G(m,m);
else
H22(i,k) = V(m)*(G(m,k)*cos(del(m)-del(k)) + B(m,k)*sin(del(m)-del(k)));
end
end
end

% H31 - Derivative of Reactive Power Injections with Angles

H31 = zeros(nqi,nbus-1);
for i = 1:nqi
m = fbus(qi(i));
for k = 1:(nbus-1)
if k+1 == m
for n = 1:nbus
H31(i,k) = H31(i,k) + V(m)* V(n)*(G(m,n)*cos(del(m)-del(n)) + B(m,n)*sin(del(m)-del(n)));
end
H31(i,k) = H31(i,k) - V(m)^2*G(m,m);
else
H31(i,k) = V(m)* V(k+1)*(-G(m,k+1)*cos(del(m)-del(k+1)) - B(m,k+1)*sin(del(m)-del(k+1)));
end
end
end
% H32 - Derivative of Reactive Power Injections with V
H32 = zeros(nqi,nbus);
for i = 1:nqi
m = fbus(qi(i));
for k = 1:(nbus)
if k == m
for n = 1:nbus
H32(i,k) = H32(i,k) + V(n)*(G(m,n)*sin(del(m)-del(n)) - B(m,n)*cos(del(m)-del(n)));
end
H32(i,k) = H32(i,k) - V(m)*B(m,m);
else
H32(i,k) = V(m)*(G(m,k)*sin(del(m)-del(k)) - B(m,k)*cos(del(m)-del(k)));
end
end
end

% H41 - Derivative of Real Power Flows with Angles

H41 = zeros(npf,nbus-1);
for i = 1:npf
m = fbus(pf(i));
n = tbus(pf(i));
for k = 1:(nbus-1)
if k+1 == m
H41(i,k) = V(m)* V(n)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n)));
else if k+1 == n
H41(i,k) = -V(m)* V(n)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n)));
else
H41(i,k) = 0;
end
end
end
end
% H42 - Derivative of Real Power Flows with V
H42 = zeros(npf,nbus);
for i = 1:npf
m = fbus(pf(i));
n = tbus(pf(i));
for k = 1:nbus
if k == m
H42(i,k) = -V(n)*(-G(m,n)*cos(del(m)-del(n)) - B(m,n)*sin(del(m)-del(n))) - 2*G(m,n)*V(m);
else if k == n
H42(i,k) = -V(m)*(-G(m,n)*cos(del(m)-del(n)) - B(m,n)*sin(del(m)-del(n)));
else
H42(i,k) = 0;
end
end
end
end

% H51 - Derivative of Reactive Power Flows with Angles

H51 = zeros(nqf,nbus-1);
for i = 1:nqf
m = fbus(qf(i));
n = tbus(qf(i));
for k = 1:(nbus-1)
if k+1 == m
H51(i,k) = -V(m)* V(n)*(-G(m,n)*cos(del(m)-del(n)) - B(m,n)*sin(del(m)-del(n)));
else if k+1 == n
H51(i,k) = V(m)* V(n)*(-G(m,n)*cos(del(m)-del(n)) - B(m,n)*sin(del(m)-del(n)));
else
H51(i,k) = 0;
end
end
end
end
 % H52 - Derivative of Reactive Power Flows with V..
H52 = zeros(nqf,nbus);
for i = 1:nqf
m = fbus(qf(i));
n = tbus(qf(i));
for k = 1:nbus
if k == m
H52(i,k) = -V(n)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n))) - 2*V(m)*(-
B(m,n)+ bpq(m,n));
else if k == n
H52(i,k) = -V(m)*(-G(m,n)*sin(del(m)-del(n)) + B(m,n)*cos(del(m)-del(n)));
else
H52(i,k) = 0;
end
end
end
end

% Measurement Jacobian, H..

H = [H11 H12; H21 H22; H31 H32; H41 H42; H51 H52];

% Gain Matrix, Gm..

Gm = H'*inv(Ri)*H;

%Objective Function
J = sum(inv(Ri)*r.^2);

% State Vector
dE = inv(Gm)*(H'*inv(Ri)*r);
E = E + dE;
del(2:end) = E(1:nbus-1);
V = E(nbus:end);
iter = iter + 1;
tol = max(abs(dE));
end
Display Results:
Del = 180/pi*del; --- State Estimation ---
E2 = [V Del]; % Bus Voltages and angles.. ----------------------------
disp('-------- State Estimation ------------------');
disp('--------------------------'); | Bus | V | Angle |
disp('| Bus |
disp('| No | pu
V | Angle | ');
| Degree | ');
| No | pu | Degree |
disp('--------------------------'); --------------------------
for m = 1:n
fprintf('%4g', m); fprintf(' %8.4f', V(m)); 1 1.0068 0.0000
fprintf(' %8.4f', Del(m)); fprintf('\n'); 2 0.9899 -5.5265
end
disp('---------------------------------------------'); 3 0.9518 -14.2039
4 0.9579 -11.4146
5 0.9615 -9.7583
You should get this in the console: 6 1.0185 -16.0798
7 0.9919 -14.7510
8 1.0287 -14.7500
9 0.9763 -16.5125
10 0.9758 -16.7476
11 0.9932 -16.5397
12 1.0009 -17.0203
13 0.9940 -17.0583
------------------------------
Appendix III: Three-Phase BCSE
Matlab Code Using WLS Method
with Power Flow Constraints
Case Study:
Three-Phase BCSE using WLS method using IEEE 13 bus
test feeder:
IEEE 13 Bus data example:
Bus V_0 V_1 V_2 d_0 d_1 d_2
1 1.0689 1.0569 1.0851 0 -120 -120
2 1.021 1.042 1.0174 -2.49 -121.72 117.83
3 1.018 1.0401 1.0418 -2.56 -121.77 117.82
4 0.994 1.0218 0.996 -3.23 -122.22 117.34
5 0 1.0311 1.0134 0 -121.9 117.86
6 0 1.0329 1.0115 0 -121.9 117.86
7 0.99 1.0529 0.9778 -5.3 -122.34 116.02
8 -0.9835 1.0553 0.9758 -5.56 -122.52 116.03
9 0 0 0.9738 0 0 115.78
10 0.9881 0 0.9758 -5.32 0 115.92
11 0.9825 0 0 -5.25 0 0
12 0.99 1.0529 0.9778 -5.3 -122.34 116.02
Importing data from file

BUS=importdata(’13bus.txt’,delimiterIn,headerlinesIn); %import bus data

BRANCH=importdata(’13branch.txt’,delimiterIn,headerlinesIn); %import line data
BUSM=importdata(’13busM.txt’,delimiterIn,headerlinesIn); %import measurement for a concrete network

bus=BUS.data(:,1);

v_0=BUS.data(:,2); v_1=BUS.data(:,3); v_2=BUS.data(:,4);

dd_0=BUS.data(:,5)*pi/180; dd_1=BUS.data(:,6)*pi/180;dd_2=BUS.data(:,7)*pi/180;

Pgen_0=BUS.data(:,8); Pgen_1=BUS.data(:,9); Pgen_2=BUS.data(:,10);

Qgen_0=BUS.data(:,11); Qgen_1=BUS.data(:,12); Qgen_2=BUS.data(:,13);

branch=BRANCH.data(:,1);
from=BRANCH.data(:,2); to=BRANCH.data(:,3); %formation of R,X matrix

r_00=BRANCH.data(:,4); r_01=BRANCH.data(:,5); r_02=BRANCH.data(:,6);

r_10=BRANCH.data(:,7); r_11=BRANCH.data(:,8); r_12=BRANCH.data(:,9);

r_20=BRANCH.data(:,10); r_21=BRANCH.data(:,11); r_22=BRANCH.data(:,12);

x_00=BRANCH.data(:,13); x_01=BRANCH.data(:,14); x_02=BRANCH.data(:,15); x_10=BRANCH.data(:,16);

x_11=BRANCH.data(:,17); x_12=BRANCH.data(:,18); x_20=BRANCH.data(:,19); x_21=BRANCH.data(:,20);
x_22=BRANCH.data(:,21); PhaseA=BRANCH.data(:,22); PhaseB=BRANCH.data(:,23); PhaseC=BRANCH.data(:,24);

Trans=BRANCH.data(:,25); ft=BRANCH.data(:,26)*0.000189393939; busM=BUSM.data(:,1); Pload_0m=BUSM.data(:,2);

Pload_1m=BUSM.data(:,3); Pload_2m=BUSM.data(:,4); Qload_0m=BUSM.data(:,5); Qload_1m=BUSM.data(:,6); Qload_2m=BUSM.data(:,7);
Declaration of symbolic variable
Ppq_0=sym(’Ppq_0’,[1 length(branch)]); % create three 1×length(branch) vectors of
symbolic variables.
Ppq_1=sym(’Ppq_1’,[1 length(branch)]);
Ppq_2=sym(’Ppq_2’,[1 length(branch)]);
PL_0=sym(’PL_0’,[1 length(bus)]);
PL_1=sym(’PL_1’,[1 length(bus)]);
PL_2=sym(’PL_2’,[1 length(bus)]);

AN=subs(A,X,Xlambda(1:length(X))); %Evaluation of symbolic variable

BN=subs(B,X,Xlambda(1:length(X)));

PL_0m=nonzeros(Pload_0m/Sbase1P); %vector P L_0m will contain active loads just for

buses which have defined load in phase zero
PL_1m=nonzeros(Pload_1m/Sbase1P);
PL_2m=nonzeros(Pload_2m/Sbase1P);
PL_0m=nonzeros(Qload_0m/Sbase1P);
QL_1m=nonzeros(Qload_1m/Sbase1P);
QL_2m=nonzeros(Qload_2m/Sbase1P);
Slack bus voltage and admittance matrix
V_01=BUS.data(1,2); % Voltage and angle for slack bus 1
V_11=BUS.data(1,3);
V_21=BUS.data(1,4);

d_01=BUS.data(1,5)*pi/180;

d_11=BUS.data(1,6)*pi/180;

d_21=BUS.data(1,7)*pi/180;

for k=1:length(branch)

Zpu1=ybase*ft(k)*[r_00(k)+1i*x_00(k)r_01(k)+1i*x_01(k)
r_02(k)+1i*x_02(k)];
Zpu2=ybase*ft(k)*[r_10(k)+1i*x_10(k)r_11(k)+1i*x_11(k)
r_12(k)+1i*x_12(k)];
Zpu3=ybase*ft(k)*[r_20(k)+1i*x_20(k)r_21(k)+1i*x_21(k)
r_22(k)+1i*x_22(k)];
Zpu=[Zpu1(Zpu1~=0);Zpu2(Zpu2~=0);Zpu3(Zpu3~=0)];ypu=inv(Zpu);

Y=abs(ypu); % Admittance matrix

thetapp=atan2(imag(ypu),real(ypu)); % Phase angle Theta
thetapq=atan2(-imag(ypu),-real(ypu));

end
Power Balance Equality constraints
for n=2:Nn
sumP_0=0;
for k=1:length(branch)
if ismember(n,from(k))&PhaseA(k)~=0
sumP_0=sumP_0+Ppq_0(k);
elseif ismember(n,to(k))&PhaseA(k)~=0
sumP_0=sumP_0+Pqp_0(k);
end;
end;

sumP_0=Pgen_0(n)-Pload_0-sumP_0;

end

where active and reactive power flows from node p to node q for phase
zero is given by:
for k=1:length(branch)
for p=1:Nn
for q=1:Nn
if ismember(p,from(k))&ismember(q,to(k))&PhaseA(k)~=0
CPpq_0=Ppq_0(k)-(V_0(p)_V_0(p)_Y(1,1)_cos(d_0(p)-d_0(p)-thetapp(1,1))+V_0(p)_V_0(q)_Y(1,1)_cos(d_0(p)-d_0(q)-
thetapq(1,1))+V_0(p)_V_1(p)_Y(1,2)_cos(d_0(p)-d_1(p)-thetapp(1,2))+V_0(p)_V_1(q)_Y(1,2)_cos(d_0(p)-d_1(q)-
thetapq(1,2))+V_0(p)_V_2(p)_Y(1,3)_cos(d_0(p)-d_2(p)-thetapp(1,3))+V_0(p)_V_2(q)_Y(1,3)_cos(d_0(p)-d_2(q)-
thetapq(1,3)));

elseif PhaseA(k)==0
CPpq_0=[];end;end;
end
end
Definition of all matrix
%Matrix of equality constraints
Con=[Eq1 Eq2];

%Matrix of variables
X=[V_0(2:length(V_0)) V_1(2:length(V_1)) V_2(2:length(V_2))
d_0(2:length(d_0)) d_1(2:length(d_1)) d_2(2:length(d_2))
Ppq Qpq PL_0P L_1 PL_2 QL_0 QL_1 QL_2];

%Matrix C, Jacobian Matrix

C=jacobian(Con,X);

%Matrix delta Z(k) , derivative of Objective

dzk=gradient(Objec,X);

%Matrix ck, the conjugate of Z

ck=Con.’;

%Matrix W, diagonal matrix of weights associated with each measurement.

W=diag(W1);

%Matrix A(leftside )AxX=B

A=[W C';C zeros(length(Con),length(Con))];

%Matrix B(rightside) AxX=B

B=[-W*dzk; -ck];
Assign Initial Values

% initial values for buses

v_0=BUS.data(:,2);
v_1=BUS.data(:,3);
v_2=BUS.data(:,4);
dd_0=BUS.data(:,5)_pi/180;
dd_1=BUS.data(:,6)_pi/180;
dd_2=BUS.data(:,7)_pi/180;
Pload_0m=BUSM.data(:,2);
Pload_1m=BUSM.data(:,3);
Pload_2m=BUSM.data(:,4);
Qload_0m=BUSM.data(:,5);
Qload_1m=BUSM.data(:,6);
Qload_2m=BUSM.data(:,7);
L=ones(1,length(Con));
% initial values for variables
Xlambda=[V_0(2:length(V_0))V_1(2:length(V_1))V_
2(2:length(V_2))
d_0(2:length(d_0)) d_1(2:length(d_1))
d_2(2:length(d_2))
Ppq Qpq PL_0 PL_1 PL_2 QL_0 QL_1 QL_2-L];
Gauss-Newton algorithm
% tolerance (stopping criteria) while(eps<epsObtained)

eps=1e−5; %in A and B substitute variables in X with

first length(X) values in Xlambda
%tolerance calculated in loop
AN=subs(A,X,Xlambda(1:length(X)));
BN=subs(B,X,Xlambda(1:length(X)));
epsObtained=1;
% calculation of numbers
%number of iterations ANE=eval(AN);
BNE=eval(BN);
iteration = 0;
% check convergence for constraints
absB=abs(BNE(length(X)+1:length(BNE)));

Display Results epsObtained=max(absB);

%calculation of dX=inv(A)_B
%This will display total number of dX=inv(ANE)_BNE;
iterations, achieved tolerance and
estimated state vector values. %update new values X(k+1)=X(k)+dX (update just
variables,not lambda)
iteration Xlambda=Xlambda(1:length(X))'+dX(1:length(X));
epsObtained Xlambda=Xlambda';
Xlambda(1:length(X))’ iteration=iteration+1; % number of iterations
end
Appendix IV: Branch Current
based State Estimation Using WLS
and Artificial Neural Network
%load error V_true_mag_total = [];
w=3; V_true_phase_total =[]; %impedance
P_start=[0.2,0.2,0.2,0.2,0.2]; load('line_impedance_1.mat');
%for w = 10:10:60 error_voltage_mag = []; from_bus = Lineimpedance(:,1);
test_error_current_real = []; error_voltage_phase = []; to_bus = Lineimpedance(:,2);
test_error_current_img = [];
test_error_current_mag = []; error_voltage_real = []; temp = from_bus(~isnan(from_bus));
test_error_current_phase = []; error_voltage_img = []; from_bus = temp;
test_error_voltage_mag = []; test = []; temp = to_bus(~isnan(to_bus));
test_error_voltage_phase = []; record_time = []; to_bus = temp;
record_x_true_full = []; clear temp;
record_z_wls = []; for q = 1:1
tic bus_list=
P_load_est = []; union(from_bus,to_bus);
Q_load_est = []; error_percen_full = []; bus_num = length(bus_list);
record_final_residual = []; error_current_mag = []; line_num = length(from_bus);
record_test = []; error_current_phase = [];
error_current_real = []; Z_pu =
p_value = []; error_current_img = []; [Lineimpedance(:,3)+1j*Lineimped
ance(:,6),Lineimpedance(:,4)+1j*
phase_est_total =[]; phase_est_final =[]; Lineimpedance(:,7),Lineimpedance
mag_est_total = []; mag_est_final = []; (:,5)+1j*Lineimpedance(:,8)];

phase_true_total =[]; % Base Selection phase_no_line = cell(1,3);

mag_true_total = []; S_base = 1000/3; phase_no_bus = cell(1,3);
V_base = 13.8/sqrt(3); z_pu = cell(1,line_num);
V_result_mag_total = []; I_base = S_base/(V_base); unbalance = cell(1,line_num);
V_result_phase_total =[]; Z_base = V_base*1000/I_base;
time_point = q;
for i = 1:line_num
z_pu{i} = Z_pu(3*(i-1)+1:3*i,1:3); I=I/I_base;
unbalance{i}=find(any(z_pu{i})==0); x_true = cell(1,3);
for j=1:3 x_ture_mag = cell(1,3);
if nnz(ismember(unbalance{i},j))>0 x_ture_phase = cell(1,3);
phase_no_line{j}=[phase_no_line{j};i];
phase_no_bus{j}=[phase_no_bus{j};to_bus(i)]; for i = 1:3
end I_r(:,i)=real(I(:,i));
end I_i(:,i)=imag(I(:,i));
end x_true{i} = [I_r(:,i);I_i(:,i)];
x_true_real{i} = I_r(:,i);
%real current x_true_img{i} = I_i(:,i);
load('allcurrents_line_head_end.mat'); x_true{i}(phase_no_line{i})=[];
line_current = []; x_true{i}(phase_no_line{i}+line_num-
for i = 1:2:34 length(phase_no_line{i}))=[];
line_current = x_true_real{i}(phase_no_line{i})=[];
[line_current;allcurrents_line_head_end(time_point,6*(i-
1)+1:6*i)]; x_true_real{i}(phase_no_line{i}+line_num-
end length(phase_no_line{i}))=[];
x_true_img{i}(phase_no_line{i})=[];

temp_i = line_current; x_true_img{i}(phase_no_line{i}+line_num-

length(phase_no_line{i}))=[];
for i = 1:line_num end

I(i,:)=[(temp_i(i,1)+1i*temp_i(i,2)),(temp_i(i,3)+1i*temp
_i(i,4)),(temp_i(i,5)+1i*temp_i(i,6))];
end
 for i = 1:18
x_true_full=[x_true{1};x_true{2};x_true{3}]
V_true_real(i,:) = [V_true(6*(i-1)+1),V_true(6*(i-
;
1)+3),V_true(6*(i-1)+5)];
x_true_real_full=[x_true_real{1};x_true_rea
V_true_img(i,:) = [V_true(6*(i-1)+2),V_true(6*(i-
l{2};x_true_real{3}];
1)+4),V_true(6*i)];
x_true_img_full=[x_true_img{1};x_true_img{2
end
};x_true_img{3}];
[V_true_phase,V_true_mag] =
cart2pol(V_true_real,V_true_img);
record_x_true_full =
[record_x_true_full,x_true_full]; %Load data
temp = xlsread('BUS_NUMBER.xlsx');
%all meter real measurement temp = temp(:,3:5);
z_1 = I(1,:); temp(1,:)=[];
z_1 = [real(z_1);imag(z_1)]; load_pu = [];

% Substation Voltage data load('load_data.mat')

load('voltage_data.mat');
V_real = zeros(bus_num,3); for i = 1:length(temp)
V =
allvoltages_rectangular_coordinates(time_poi load_pu(i,:)=[((P_load(time_point,i)+1j*Q_load(time
nt,1:6)/(V_base*1000); _point,i))*temp(i,1)/sum(temp(i,:))),((P_load(time_
point,i)+1j*Q_load(time_point,i))*temp(i,2)/sum(tem
%create true voltage data p(i,:))),((P_load(time_point,i)+1j*Q_load(time_poin
V_true = t,i))*temp(i,3)/sum(temp(i,:)))];
allvoltages_rectangular_coordinates(time_poi end
nt,:)/(V_base*1000);
load_pu = load_pu/S_base;
V_true_real = zeros(18,3); load_pu_P=real(load_pu);
V_true_img = zeros(18,3); load_pu_Q=imag(load_pu);
residual_real=[]; % Build h
residual_img = []; h=zeros(2*bus_num-2,2*line_num);
for i=2:bus_num
load('end_bus_0load.mat') for j=1:line_num
if to_bus(j)==i
% run x times to stable random Gaussian error h(i-1,j)=1;
for number = 1:1 h(i+bus_num-2,j+line_num)=1;
V_ref_1_real = normrnd(V(1,1),abs(V(1,1))*3/(3*100)); else
V_ref_1_imag = normrnd(V(1,2),abs(V(1,2))*3/(3*100)); if from_bus(j)==i
V_ref_2_real = normrnd(V(1,3),abs(V(1,3))*3/(3*100)); h(i-1,j)=-1;
V_ref_2_imag = normrnd(V(1,4),abs(V(1,4))*3/(3*100)); h(i+bus_num-2,j+line_num)=-1;
V_ref_3_real = normrnd(V(1,5),abs(V(1,5))*3/(3*100)); end
V_ref_3_imag = normrnd(V(1,6),abs(V(1,6))*3/(3*100)); end
end
V_real(1,:)=[V_ref_1_real+1i*V_ref_1_imag end
V_ref_2_real+1i*V_ref_2_imag
V_ref_3_real+1i*V_ref_3_imag]; weight = [1*10^-5*ones(34,1);1*10^-6*ones(2,1)];
V_ref = V_real(1,:); temp_weight = repmat(weight,3);
temp_weight(:,2:3)=[];
%set some error for initial guess current value convergence = [];
I = 0.2*I;
sigma_pse_P=abs(load_pu_P)*w/(3*100);
%Forward Sweep
for i = 1:17 sigma_pse_Q=abs(load_pu_Q)*w/(3*100);
V_real(to_bus(i),:) = V_real(from_bus(i),:)-
(Z_pu(3*(i-1)+1:3*i,:)*I(i,:).').';
end
 load_pu_P_n=normrnd(load_pu_P,sigma_pse_P); h_wls=cell(1,3);
load_pu_Q_n=normrnd(load_pu_Q,sigma_pse_Q); for i=1:3
load_pu_n=(load_pu_P_n+1j*load_pu_Q_n); % P & Q
h1=zeros(2,2*line_num);
h1(1,1)=1;
%real; add error for real measurement h1(2,1+line_num)=1;
sigma_real=abs(z_1)*3/(3*100);
z_1_n=normrnd(z_1,sigma_real); h2=zeros(2,2*line_num);
h2(1,2)=1;
I_r=zeros(line_num,3); h2(2,2+line_num)=1;
I_i=zeros(line_num,3);
x_est_old=cell(1,3); h3=zeros(2,2*line_num);
x_est_old_full=[]; h3(1,3)=1;
for i=1:3 h3(2,3+line_num)=1;
I_r(:,i)=real(I(:,i));
I_i(:,i)=imag(I(:,i)); h4=zeros(2,2*line_num);
x_est_old{i}=[I_r(:,i);I_i(:,i)]; h4(1,4)=1;
x_est_old{i}(phase_no_line{i})=[]; h4(2,4+line_num)=1;
x_est_old{i}(phase_no_line{i}+line_num-
length(phase_no_line{i}))=[]; h5=zeros(2,2*line_num);
x_est_old_full=[x_est_old_full;x_est_old{i}]; h5(1,5)=1;
end h5(2,5+line_num)=1;
V_real(1,:)=[];
pse=conj(load_pu_n./V_real); h6=zeros(2,2*line_num);
h6(1,6)=1;
z=zeros(2*bus_num-2,3); h6(2,6+line_num)=1;
for i=1:3
z(:,i)=[real(pse(:,i));imag(pse(:,i))];
end
h7=zeros(2,2*line_num); h14=zeros(2,2*line_num); % Build full H
h7(1,7)=1; h14(1,14)=1; [m1,n1]=size(h_wls{1});
h7(2,7+line_num)=1; h14(2,14+line_num)=1; [m2,n2]=size(h_wls{2});
[m3,n3]=size(h_wls{3});
h8=zeros(2,2*line_num); h15=zeros(2,2*line_num);
h8(1,8)=1; h15(1,15)=1; h_wls_full=zeros(m1+m2+m3,n1+n2
h8(2,8+line_num)=1; h15(2,15+line_num)=1; +n3);

h9=zeros(2,2*line_num); h16=zeros(2,2*line_num); h_wls_full(1:m1,1:n1)=h_wls{1};

h9(1,9)=1; h16(1,16)=1;
h9(2,9+line_num)=1; h16(2,16+line_num)=1; h_wls_full(m1+1:m1+m2,n1+1:n1+n
2)=h_wls{2};
h10=zeros(2,2*line_num); h17=zeros(2,2*line_num);
h10(1,10)=1; h17(1,17)=1; h_wls_full(m1+m2+1:m1+m2+m3,n1+
h10(2,10+line_num)=1; h17(2,17+line_num)=1; n2+1:n1+n2+n3)=h_wls{3};
h_wls{i}=[h;h1];
h11=zeros(2,2*line_num); z_wls=cell(1,3);
h11(1,11)=1; h_wls{i}(:,phase_no_line{i})=[]; z_wls_full=[];
h11(2,11+line_num)=1; h_wls{i}(:,phase_no_line{i}+line_num- for i=1:3
length(phase_no_line{i}))=[];
h12=zeros(2,2*line_num); end z_wls{i}=[z(:,i);z_1_n(:,i)];
h12(1,12)=1;
h12(2,12+line_num)=1; z_wls_full=[z_wls_full;z_wls{i}
];
h13=zeros(2,2*line_num); end
h13(1,13)=1;
h13(2,13+line_num)=1;
h_zero_index=all(h_wls_full==0, 2); for phase=1:3
h_wls_full(h_zero_index,:)=[]; correct_factor=0;
z_wls_full(h_zero_index)=[]; for i=1:line_num
if
record_z_wls = [record_z_wls,z_wls_full]; nnz(ismember(phase_no_line{phase},i))==0
I_new(i,phase)=x_est_temp{phase}(i-
%SE loop correct_factor)+1i*x_est_temp{phase}(i+line_num-
iter = 0; length(phase_no_line{phase})-correct_factor);
while 1 else
Rinv_full=eye(length(z_wls_full)); I_new(i,phase)=0;
for i=1:length(z_wls_full) correct_factor=correct_factor+1;
Rinv_full(i,i)=1/(temp_weight(i)^2); end
end end
end
Gain=h_wls_full.'*Rinv_full*h_wls_full;
beta=h_wls_full.'*Rinv_full*z_wls_full; V_new=zeros(bus_num,3);
x_est_full=Gain\beta; V_new(1,:)=[V_ref_1_real+1i*V_ref_1_imag
x_est_temp=cell(1,3); V_ref_2_real+1i*V_ref_2_imag
x_est_temp{1}=x_est_full(1:n1); V_ref_3_real+1i*V_ref_3_imag];
x_est_temp{2}=x_est_full(n1+1:n1+n2);
x_est_temp{3}=x_est_full(n1+n2+1:n1+n2+n3); for i = 1:17
I_new=zeros(line_num,3); V_new(to_bus(i),:) = V_new(from_bus(i),:)-
(Z_pu(3*(i-1)+1:3*i,:)*I_new(i,:).').';
end
convergence=[convergence;max(x_est_full-
x_diff = x_true_full-x_est_full;
x_est_old_full)];
err_wls = sqrt(mean((x_true_full-x_est_full).^2));
time = toc
if max(x_est_full-x_est_old_full)<1e-6
V_real=V_new;
record_time = [record_time;time];
I=I_new;
x_est_full_real =
break;
[x_est_full(1:17);x_est_full(35:51);x_est_full(69:85)];
end
x_est_full_img =
[x_est_full(18:34);x_est_full(52:68);x_est_full(86:102)];
% update z_wls value
iter=iter+1;
x_true_full_real =
x_est_old_full=x_est_full;
[x_true_full(1:17);x_true_full(35:51);x_true_full(69:85)]
V_real=V_new;
;
I=I_new;
x_true_full_img =
V_real(1,:)=[];
[x_true_full(18:34);x_true_full(52:68);x_true_full(86:102
pse=conj(load_pu_n./V_real);
)];
z=zeros(2*bus_num-2,3);
% transfer to phase & magnitude
for i=1:3
[phase_est,mag_est]=
z(:,i)=[real(pse(:,i));imag(pse(:,i))];
cart2pol(x_est_full_real,x_est_full_img);
end
[phase_true,mag_true]=
cart2pol(x_true_real_full,x_true_img_full);
z_wls=cell(1,3);
z_wls_full=[];
%forward for bus Voltage
for i=1:3
current_temp = mag_est .* cos(phase_est) +
1i.*mag_est .* sin(phase_est);
z_wls{i}=[z(:,i);z_1_n(:,i)];
current_result = [];
z_wls_full=[z_wls_full;z_wls{i}];
end
z_wls_full(h_zero_index)=[];
end
 for i = 1:17
current_result = %votlage error
[current_result;current_temp(i),current_temp(i+17),curren error_voltage_mag =
t_temp(i+34)]; [error_voltage_mag;abs(mean((V_result_mag-
end V_true_mag)/mean(V_true_mag)))*100];
error_voltage_phase =
V_result=zeros(bus_num,3); [error_voltage_phase;abs(mean((V_result_phase-
V_result(1,:)=[V_ref_1_real+1i*V_ref_1_imag V_true_phase)/mean(V_true_phase)))*100];
V_ref_2_real+1i*V_ref_2_imag
V_ref_3_real+1i*V_ref_3_imag]; end

for i = 1:17 phase_est_total = [phase_est_total;phase_est];

V_result(to_bus(i),:) = V_result(from_bus(i),:)- mag_est_total = [mag_est_total;mag_est];
(Z_pu(3*(i-1)+1:3*i,:)*current_result(i,:).').'; %
end phase_true_total =
[phase_true_total;phase_true];
V_result_real = real(V_result); mag_true_total=[mag_true_total;mag_true];
V_result_img = imag(V_result);
[V_result_phase,V_result_mag] =
cart2pol(V_result_real,V_result_img); V_result_mag_total =
[V_result_mag_total;V_result_mag];
error_voltage_real = V_result_phase_total =
[error_voltage_real;abs(mean((V_result_real- [V_result_phase_total;V_result_phase];
V_true_real)/mean(V_true_real)))*100];
error_voltage_img = V_true_mag_total = [V_true_mag_total;V_true_mag];
[error_voltage_img;abs(mean((V_result_img- V_true_phase_total =
V_true_img)/mean(V_true_img)))*100]; [V_true_phase_total;V_true_phase];
end
% ANN-based estimator %calculate error
MAPE = [];
temp_training_data = record_z_wls'; for i = 1:200
temp_training_output = record_x_true_full'; MAPE = [MAPE;mean((abs(testing_label(i,:))-
abs(estimate_testing_label(i,:)))./abs(testing_label(i,:)
%80% for training, 20% for testing, random pick ))*100];
test_index = randperm(length(record_x_true_full)); end
training_data =
temp_training_data(test_index(1:800),:); MAPE = [];
training_label = for i = 1:200
temp_training_output(test_index(1:800),:); MAPE = [MAPE;mean(abs(testing_label(i,:)-
testing_data = estimate_testing_label(i,:))./abs(testing_label(i,:)))*10
temp_training_data(test_index(801:1000),:); 0];
testing_label = end
temp_training_output(test_index(801:1000),:);

%hyperparameter, take care out-of-memory problem.

net1_fine_tune = feedforwardnet([10,10,10]);
net1_fine_tune =
train(net1_fine_tune,training_data',training_label');

estimate_testing_label =
net1_fine_tune(testing_data');
estimate_testing_label = estimate_testing_label';