State Estimation
DIgSILENT PowerFactory *
Abstract The single line diagram is shown in Figure 1.
This paper describes the set-up of a power net-
work model for state estimation, which is part of
2 Model Parameters
the PowerFactory examples. The parameters of
the individual elements as well as the assignment The following subsections describe the parameters of
of measurement values for input in the PowerFac- the elements as used for balanced load flow calcula-
tory network model are explained. Results for the tion and state estimation.
state estimation calculation in PowerFactory are
presented. The model is set-up to enable both State Estimation
and balanced Load Flow Calculation.
1 Network Model
2.1 Loads
The network model, which is used in the state estima- During the load flow calculation, the loads of the state
tion example, represents a small distribution system. estimation example have constant active and reactive
The nominal voltages of the power system are 115 kV, power demand and are not voltage-dependent. This
20 kV and 0,4 kV, the nominal frequency is 50 Hz. The is achieved by disabling the load option Consider Volt-
system is supplied by one feeding substation with two age Dependency of Loads in the PowerFactory Load
parallel HV/MV transformers. One of the transformers Flow Calculation command.
is switched-off and considered to be on stand-by.
Load data (active power 𝑃 and reactive power 𝑄) is
The MV grid consists of 5 buses (nodes), 5 MV/LV listed in Table 1.
transformers, 3 loads and 5 lines. A radial MV feeder
starts at line MV-K1. The lines MV-K4 and MV-K7 sup- Depending on the study case, the active power 𝑃 and
ply a low voltage grid with 9 nodes, 9 LV cables and reactive power 𝑄 of the loads are to be estimated. On
5 loads. The network is meshed between the busbars the State Estimation page of the input dialog of the
Central and RW-South-MV. loads, there are tick boxes that determine whether 𝑃
and/or 𝑄 shall be estimated.
* DIgSILENT GmbH, Heinrich-Hertz-Str. 9, 72810 Gomaringen, Germany, www.digsilent.de
STATE ESTIMATION
Figure 1: Single line diagram of the State estimation example system
2.2 External Grid The 𝑉 measurements are set out-of-service.
The actual measurement values as well as their sta-
The external grid “Ext. Netw.” is used as a slack net-
tus (“in service” or “out of service”) are different in the
work injection with voltage 1.0 p.u. and 0 degrees in
3 study cases.
the Load Flow page. Other generating units are not
present in the state estimation system. In study case “1 - Undisturbed Measurements”,
the scenario “Base Scenario (undisturbed measure-
For state estimation, the active power 𝑃 and the reac-
ments)” is activated. All 𝐼, 𝑃 and 𝑄 measurements
tive power 𝑄 at the external grid are determined from
are active and their values are equal to the load flow
the measurement values in the model.
results at the respective branches, nodes and loads
when the actual tap position of transformer “NT1” is
set to 4.
2.3 Measurement values
Activating study case “2 - Disturbed Measurements”,
Measurement objects for active power 𝑃 , reactive the scenario “Scenario with disturbed measurements”
power 𝑄, voltage magnitude 𝑉 as well as for active becomes active. All 𝐼, 𝑃 and 𝑄 measurement are ac-
and reactive part of the current “I” are defined in each tive, but their values are noisy, reflecting typical mea-
connection point of all MV and LV lines and transform- surement inaccuracy.
ers, MV side of HV/LV-transformers and loads. In total,
50 𝑃 measurements, 50 𝑄 measurements, 50 𝐼 mea- In the scenario “Unobservable Scenario” that is acti-
surements and 50 𝑉 measurements are given. The vated together with study case “3 - Unobservable Net-
𝐼 measurements and the pairs of 𝑃 and 𝑄 measure- work”, several measurements have been put out of
ments are redundant at each location. service to model a set-up in which no complete cov-
erage of measurement points is given. As a conse-
Rev.1 2
STATE ESTIMATION
quence, not all load power values can be determined all three study cases. They are depicted in Figure 2
from the remaining active measurements. Moreover, and additionally provided in Appendix B.
some of the measurements have been manipulated
to drastically differ from the load flow results at the re-
spective branches, which is supposed to model severe
measurement errors or misconfigurations. 4 State estimation
The state estimation is examined by executing the
2.4 Estimated States State Estimation command ( ). Optionally, various
plausibility checks can be performed to exclude obvi-
Beside the location and values of measurements, the ously incorrect measurements from the state estima-
estimated states need to be defined for a state estima- tion. Furthermore, an observability check can be run
tion calculation. In the state estimation example, the to investigate if the available measurement information
estimated states are the parameters active power 𝑃 is sufficient to determine the estimated states.
and reactive power 𝑄 for all loads. The state estima-
tion optimisation method will then determine their val- The state estimation is performed for different sets of
ues such that the resulting load flow optimally matches measurement values. This illustrates on the one hand
the measurement values. The respective flags are set the influence of measurement deviation on the esti-
in the input dialog of the loads on the State Estimation mated load values. On the other hand the influence of
page. the number and distribution of measurements on the
network observability is shown.
The tap position of the transformers “NT1, NT2, LV-
T1, LV-T2” and “LV-T5” can be set as states to be es-
timated on the page State Estimation in the input dia- 4.1 Undisturbed measurements
log of the transformers. These states are considered
in study cases “1 - Undisturbed Measurements” and
In this study case the measurement values reflect the
“2 - Disturbed Measurements”.
load flow results at each branch for the tap position
The estimated states vary in the three study cases. of transformer “NT1” set to 4. In the scenario the tap
While study case “3 - Unobservable Network” only in- position was changed to neutral position 9 to illustrate
vestigates the 8 active power 𝑃 values of the loads. the influence of changes in the network model on the
The active variation “Unobservable Network” stores state estimation result.
this modification of load and transformer data.
The estimated states are the active power values of all
loads.
3 Load Flow Results The state estimation calculation is executed. None
of the options for additional plausibility checks are
selected. The check for observable regions is per-
The steady-state load flow is examined by executing formed. In unobservable regions, the 𝑃 and 𝑄 values
the Load Flow Calculation command ( ). The results of the loads defined as input for load flow calculation
of the PowerFactory load flow calculation are equal for would be used.
Rev.1 3
STATE ESTIMATION
Figure 2: Results of the load flow calculation
All measurements are valid, only 32 of the 150 mea- object.
surements are relevant for the calculation of the ob-
served load states. The rest are redundant. The estimated states are the active power and reac-
tive power values of all loads. The tap positions of
The resulting 𝑃 values for the loads are shown in Ta- the transformers “NT1, NT2, LV-T1, LV-T2” and “LV-
ble 4. The branch flows and node voltages computed T5” are also to be estimated. Twenty states are to be
by the state estimation calculation slightly differ from estimated in total.
the original load flow results due to the modified trans-
former tap position at “NT1”. All options for additional plausibility checks are se-
lected in the state estimation command. The plau-
The measured and calculated values are compared sibility checks of measurement values are performed
for each type of measurements. This is shown exem- before the actual state estimation optimisation. The
plarily for the active power measurements in Table 5. tests fail for 7 𝑃 measurements and 4 𝑄 measure-
There are no bad measurements detected. ments. Table 6 shows the detected inconsistencies.
The respective measurements are invalid and there-
fore excluded from the observability check and state
4.2 Impact of Measurement Errors estimation.
The check for observable regions is performed. All 20
The simulation “2 - Disturbed measurements” takes states are observable by means of the remaining valid
into account measurement errors. The state estima- measurements.
tion calculation balances those deviations by means
of a weighted least mean square optimisation. The The state estimation identifies bad measurements
weighting factors are determined from the Power Rat- whenever the calculated value differs from the mea-
ing and Accuracy Class given for each measurement sured value by more than the percentage given as the
Rev.1 4
STATE ESTIMATION
Accuracy Class. These measurements are excluded mand are omitted in this study case.
from the calculation. The observability check and the
state estimation are iteratively repeated until no more The check for observability regions is activated. The
bad measurements are found. The results of the state treatment of unobservable regions is set to the option
estimation are depicted in Figure 3. Bad measure- use only internally crated pseudo-measurements. Us-
ments are coloured in magenta and invalid measure- ing this option, in case of unobservable network ar-
ments in red. Table 7 additionally lists the bad mea- eas, PowerFactory will automatically include pseudo-
surements. measurements at locations of non-observable (load)
states. These pseudo-measurements will assume the
active and/or reactive power values from the load flow
4.3 Impact of the Number of Measure- page of the loads. However, their accuracy is very low
compared to real measurements, resulting in small
ments
weighting factors for the least mean square optimisa-
tion. In this sense, pseudo-measurements will help to
In order to simulate the impact of the number of avail- make the system artificially observable with a minimal
able measurements and their distribution in the net- impact on the solution of the state estimation.
work on the observability of the estimated states, sev-
eral measurements are switched-off in study case “3 The state estimation is performed. The result of the
- Unobservable Network”. Only 8 𝑃 measurements, 8 observability check is given in the PowerFactory out-
𝑄 measurements and 10 𝐼 measurements remain ac- put window. Only 14 of the 20 states are observable.
tive. The rest is set out-of-service. The same 20 states Table 8 lists the non-observable states of the loads.
as in subsection 4.2 are to be estimated from these Internal pseudo-measurements are defined instead by
26 measurement values. The colouring mode “Sec- PowerFactory. The state estimation identifies 4 bad 𝑃
ondary Equipment - Measurement Locations” makes measurements as listed in Table 7.
the measurements visible.
The resulting estimated load states are presented in
The plausibility checks in the state estimation com- Table 9.
Rev.1 5
STATE ESTIMATION
Figure 3: State estimation results and invalid or bad measurements for the study case “2 - Disturbed Measure-
ments”
Rev.1 6
STATE ESTIMATION
A Tables with Input Data
Table 1: Load Demand
Load Bus P [MW] Q [Mvar]
A. Lincoln Str. Railway Station 0,30 0,31
Feeder A Airport 1,00 0,48
Feeder B Airport 2,00 1,50
Industrial Area West 3,00 3,06
J. Carter Str. Railway Station 0,50 0,24
KV101 Baker Str. 0,05 0,04
KV301 Term. Strip 1 0,05 0,04
Recr. Area Recrat. 0,40 0,19
B Tables with Results of the Balanced Load Flow Calculation
Table 2: Results of bus voltages obtained with PowerFactory
U, Magnitude (line-line) u, Magnitude u, Angle (line-earth)
Name
[kV] [p.u.] [deg]
SS1 115,00 1,000 0,00
SS2 115,00 1,000 0,00
Airport 18,10 0,905 -1,56
BB1l 18,50 0,925 -1,63
BB1r 18,50 0,925 -1,63
BB2l 18,50 0,925 -1,63
BB2r 18,50 0,925 -1,63
Central 18,48 0,924 -1,65
East 18,47 0,923 -1,65
RW-South MV 18,46 0,923 -1,66
West 18,02 0,901 -1,52
Baker Str. 0,36 0,888 27,27
Central-MV 0,36 0,890 27,29
East-LV 0,36 0,888 26,85
RW-South 0,36 0,888 27,23
Railway Station 0,35 0,865 27,25
Recrat. 0,35 0,868 26,71
Rev.1 7
STATE ESTIMATION
Table 3: Results of lines obtained with PowerFactory
Loading Active Reactive Current Losses Reactive
Name
[%] Power [MW] Power [Mvar] [kA] [MW] Losses [Mvar]
K101a 4,95 0,05 0,02 0,089 0,000 0,000
K101b 1,27 0,00 -0,01 0,023 0,000 0,000
K201a 32,55 0,29 0,21 0,586 0,008 0,006
K201b 57,35 -0,52 -0,34 1,032 0,007 0,004
K201c 57,35 -0,52 -0,35 1,032 0,007 0,004
K300a 46,69 -0,45 -0,23 0,840 0,005 0,003
K300b 41,07 -0,40 -0,20 0,739 0,001 0,001
K300c 41,07 -0,40 -0,19 0,739 0,004 0,002
MV-K1 87,57 6,16 4,84 0,244 0,139 -0,080
MV-K2 47,97 3,02 2,93 0,134 0,017 -0,058
MV-K3 0,79 0,00 -0,07 0,002 0,000 -0,070
MV-K4 10,10 -0,81 -0,43 0,029 0,001 -0,109
MV-K5 5,84 0,46 0,18 0,016 0,000 -0,073
MV-K7 7,27 0,54 0,09 0,017 0,001 -0,292
Sp. Coup. 0,00 0,00 0,00 0,000 0,000 0,000
C Tables with results of the State Estimation
Table 4: Input Values and State Estimation results of Active Power 𝑃 of Loads
Input P Input Q Result P Result Q
Load Bus
[MW] [Mvar] [MW] [Mvar]
A. Lincoln Str. Railway Station 0,300 0,306 0,300 0,306
Feeder A Airport 1,000 0,484 1,000 0,484
Feeder B Airport 2,000 1,500 2,000 1,500
Industrial Area West 3,000 3,061 3,000 3,061
J. Carter Str. Railway Station 0,500 0,242 0,500 0,242
KV101 Baker Str. 0,050 0,038 0,050 0,038
KV301 Term. Strip 1 0,050 0,038 0,050 0,038
Recr. Area Recrat. 0,400 0,194 0,400 0,194
Table 5: Comparison of Input and Calculated Measurement Values for Active Power 𝑃
Measured 𝑃 Calculated 𝑃 Deviation
Measurement
[MW] [MW] [%]
Pmea (A. Lincoln Str.; Railway Station) 0,300 0,300 0,000
Pmea (Feeder A; Airport) 1,000 1,000 0,000
Pmea (Feeder B; Airport) 2,000 2,000 0,000
Pmea (Industrial Area; West) 3,000 3,000 0,000
Pmea (J. Carter Str.; Railway Station) 0,500 0,500 0,000
Pmea (K101a; Baker Str.) -0,050 -0,050 0,000
Pmea (K101a; Central-MV) 0,050 0,050 0,000
Pmea (K101b; Baker Str.) 0,000 0,000 0,000
Pmea (K101b; RW-South) 0,000 0,000 0,000
Rev.1 8
STATE ESTIMATION
Table 5: Comparison of Input and Calculated Measurement Values for Active Power 𝑃
Measured 𝑃 Calculated 𝑃 Deviation
Measurement
[MW] [MW] [%]
Pmea (K201a; Central-MV) 0,290 0,290 0,000
Pmea (K201a; Railway Station) -0,284 -0,284 0,000
Pmea (K201b; Term. Strip 3) 0,522 0,522 0,000
Pmea (K201bc; RW-South) 0,527 0,527 0,000
Pmea (K201bc; Railway Station) -0,516 -0,516 0,000
Pmea (K201c; Term. Strip 3) -0,522 -0,522 0,000
Pmea (K300; East-LV) 0,457 0,457 0,000
Pmea (K300; Recrat.) -0,400 -0,400 0,000
Pmea (K300a; Term. Strip 1) -0,454 -0,454 0,000
Pmea (K300b; Term. Strip 1) 0,404 0,404 0,000
Pmea (K300b; Term. Strip 2) -0,403 -0,403 0,000
Pmea (K300c; Term. Strip 2) 0,403 0,403 0,000
Pmea (KV101; Baker Str.) 0,050 0,050 0,000
Pmea (KV301; Term. Strip 1) 0,050 0,050 0,000
Pmea (LV-T1; Central) 0,171 0,171 0,000
Pmea (LV-T1; Central-MV) -0,170 -0,170 0,000
Pmea (LV-T2; Central) 0,171 0,171 0,000
Pmea (LV-T2; Central-MV) -0,170 -0,170 0,000
Pmea (LV-T3; East) 0,230 0,230 0,000
Pmea (LV-T3; East-LV) -0,229 -0,229 0,000
Pmea (LV-T4; East) 0,230 0,231 0,125
Pmea (LV-T4; East-LV) -0,229 -0,229 0,000
Pmea (LV-T5; RW-South MV) 0,531 0,531 0,000
Pmea (LV-T5; RW-South) -0,528 -0,528 0,000
Pmea (MV-K1; Airport) -6,013 -6,013 0,000
Pmea (MV-K1; BB2l) 6,125 6,125 0,000
Pmea (MV-K2; Airport) 3,013 3,013 0,000
Pmea (MV-K2; West) -3,000 -3,000 0,000
Pmea (MV-K3; East) 0,000 0,000 0,000
Pmea (MV-K3; West) 0,000 0,000 0,000
Pmea (MV-K4; BB2l) 0,803 0,803 0,000
Pmea (MV-K4; Central) -0,802 -0,802 0,000
Pmea (MV-K5; Central) 0,461 0,461 0,000
Pmea (MV-K5; East) -0,460 -0,460 0,000
Pmea (MV-K7; BB2r) 0,532 0,532 0,000
Pmea (MV-K7; RW-South MV) -0,531 -0,531 0,000
Pmea (NT1; BB2l) -7,460 -7,460 0,000
Pmea (Recr. Area; Recrat.) 0,400 0,400 0,000
Rev.1 9
STATE ESTIMATION
Table 6: Measurements that failed plausibility checks in study case “2 - Disturbed Measurements”
Measurement Failed Plausibility Check
Pmea (MV-K1; BB2l) Branch losses exceed nominal values
Pmea (MV-K1; Airport) Branch losses exceed nominal values
Pmea (MV-K3; East) Branch losses exceed nominal values
Pmea (MV-K3; West) Branch losses exceed nominal values
Pmea (K300b; Term. Strip 2) Negative losses on passive branches
Pmea (K300b; Term. Strip 1) Negative losses on passive branches
Pmea (NT2; BB1r) Large branch flows on open ended branches
Pmea (MV-K3; East) Large branch flows on open ended branches
Pmea (NT2; BB1r) Node sum check for P exceeds tolerance
Pmea (MV-K3; East) Node sum check for P exceeds tolerance
Table 7: Bad Measurements in study case “2 - Disturbed Measurements”
Measured 𝑃 Calculated 𝑃 Deviation
Measurement
[MW] [MW] [%]
Pmea (K201a; Railway Station) -0,199 -0,279 -8,046
Pmea (K201bc; RW-South) 0,597 0,519 -7,851
Pmea (LV-T1; Central) 0,091 0,166 7,541
Pmea (LV-T3; East-LV) -0,274 -0,220 5,399
Pmea (LV-T5; RW-South MV) 0,606 0,524 -8,132
Pmea (Recr. Area; Recrat.) 0,505 0,391 -11,379
Pmea (Sp. Coup.; Term. Strip 2) -0,050 0,000 5,000
Table 8: Non-observable States of Loads in study case “3 - Unobservable Network”
Load Busbar 𝑃 observable? 𝑄 observable?
A. Lincoln Str. Railway Station yes no
Feeder A Airport no yes
Feeder B Airport yes yes
Industrial Area West no yes
J. Carter Str. Railway Station yes no
KV101 Baker Str. yes yes
KV301 Term. Strip 1 no yes
Recr. Area Recrat. no yes
Rev.1 10
STATE ESTIMATION
Table 9: Input values and State Estimation results of Active Power 𝑃 of Loads in study case
“3 - Unobservable Network”
Input P Input Q Result P Result Q
Load Bus
[MW] [Mvar] [MW] [Mvar]
A. Lincoln Str. Railway Station 0,300 0,306 0,280 0,296
Feeder A Airport 1,000 0,484 1,003 0,419
Feeder B Airport 2,000 1,500 1,972 1,404
Industrial Area West 3,000 3,061 3,003 3,010
J. Carter Str. Railway Station 0,500 0,242 0,420 0,233
KV101 Baker Str. 0,050 0,038 0,123 -0,021
KV301 Term. Strip 1 0,050 0,038 0,050 0,016
Recr. Area Recrat. 0,400 0,194 0,400 0,187
