State Estimation at All India Level

Akhil Singhal Harish Kumar Rathour P.K. Agarwal KVS Baba S.K. Soonee
Deputy Manager Chief Manager General Manager Executive Director Chief Executive Officer
POWER SYSTEM OPERATION CORPORATION LIMITED, NEW DELHI, INDIA

Abstract — Indian power system is one of the largest n is the number of measurements.
synchronously operated Power System Networks in the It is the function ‘f’ which is minimized using Weighted
world. It has an installed capacity of about 303 GW. Least Square algorithm.
Indian power system has complex transmission network
along with many cross border connections. It is monitored
State Variables: Voltage and Angle at each node (refer
by a multi-level integrated Supervisory Control & Data fig.1) are the state variables. By estimating these two
Acquisition (SCADA)/ Energy Management System (EMS) parameters, we can calculate line flows, injection, drawls,
installed at many control centres dispersed all over India. etc.
SCADA/EMS of National control centre is integrated with
SCADA/EMS systems of five Regional control centres.
SCADA/EMS systems of Regional control centres in turn
are integrated with SCADA/EMS systems of 32 State
Fig.1: Common Bus diagram showing state variables
control centres.
Real Power, P = V1*V2*sin(θ1- θ2)
It is important to use all real-time tools to help the operator -------------------------
in maintaining the grid in a highly secure and reliable XL
manner. Successful implementation of State Estimator is Reactive Power, Q = V1 2 – V1V2 cos(θ1- θ2)
said to be critical requirement for efficient EMS. An --------------------------------
exercise was taken up by Power System Operation XL
Corporation Limited (POSOCO) with an objective of II. Use of State Estimation in Energy Management
improving the usage of State Estimation in Load Despatch
System at Load Despatch Centres
Centres in India. In this process, it has been interacting
with experts both in India and abroad with the objective of Energy Management System (EMS) is the mechanism to
learning and making continuous improvement. capture “system knowledge” and “situational awareness”
and provide some key indicators to help the Load
Successful running of State Estimation at All India level Despatcher in real-time decision making. EMS tools are
involves maintaining and tuning of a large SCADA and essential for reliable and secure operation of the grid.
Network database. Such a big database needs correction in
topology, device parameters, connectivity model, etc. This EMS applications can be divided into two parts –
paper focusses on all such related issues, its solution and a) Network Applications which includes fault analysis,
impact on State Estimator output. contingency analysis, Powerflow, optimal powerflow,
Keywords – Supervisory Control and Data Acquisition Data Validation (by State Estimation), etc.
(SCADA), National Load Despatch Center (NLDC), b) Generation Applications which includes Unit
Energy Management System (EMS), State Estimator (SE), commitment/scheduling, Reserve/Cost monitoring,
Power Network Applications (PNA). Economic Despatch, etc.
I. Introduction Energy Management System (EMS) has a system of
State Estimation was introduced by Gauss and Legendre computer-aided tools used by operators of electric utility
(around 1800). The basic idea was to “fine tune” State grids to monitor, control and optimize the performance of
Variables (Voltage and Angle at each bus) by minimizing the generation and/or transmission system. The monitor and
the sum of the residual squares. This is the well-known control functions are known as SCADA. This computer
Least Squares (LS) method. The reason for its popularity is technology is also referred to as SCADA/EMS or
easy to understand: At the time of its invention there were EMS/SCADA. The terminology EMS then excludes the
no computers, and the fact that the Least Squares estimator monitoring and control functions, but more specifically
could be computed explicitly from the data made it the only refers to the collective suite of Power Network Applications
feasible approach. State Estimation was applied to Power (PNA) and to the generation control and scheduling
Systems by Schweppe and Wildes in the late 1960s in a applications.
real-time environment. In Load Despatch Centre at National level, thousands of
The basic equations involved are – analog and digital data are acquired in real-time which may
also contain some incorrect data or some data might not be
[z] = H[x] + e getting telemetered due to various reasons such as faulty
f = Σ Wj.ej2 (j=1 to n) RTU, faulty transducers, poor state of potential-free NO/NC
Where, contacts in switching devices, communication failure, etc.
z is the vector of measured state variables.
H is measurement function coefficient matrix.
x is the vector of estimated state variables.
e is the error.
Fig.2: Non-availability of Analog data due to various reasons
Wj is the weight assigned to error ej.

1 of 6
State Estimation provides an estimated value of measurands
for all correctly telemetered/non-telemetered/incorrect data
and hence acts as a cross-check reference helping in Bad
Data detection.
Output of the State Estimator provides Voltage Magnitude
and Angle at each bus and hence gives a base-case for all
the Network Applications and some of the generation
applications to carry out various studies in order to help grid
operator in real-time operation.

Fig.5: Display at NLDC to facilitate truncation of network at 400kV

and above level
2. Network Parameters

Fig.3: Base case generation from State Estimation process

State Estimation requires correct model parameters of the
network. Incorrect parameters of various power system
III. Corrections Done to Improve State Estimator equipments impact the State Estimator output.
Solution
2.1 Correction in Line Parameters
1. Network Truncation
Incorrect line parameters cause change in impedance of
The Network Model at NLDC is vast consisting of more paths, hence different flows (MW and MVAR) than actual,
than 1800 sub-stations. Updating and maintaining (or and change in the estimated angles at the station buses.
tuning) such a large network model for State Estimation is Many cases of incorrect parameters were observed and
very difficult. In order to focus on the main network and rectified. An example is shown in Fig.6.
considering only on the major portion (say 400kV level and
above) Network Truncation is done.

(a) Network before truncation

Fig.6: Impact of different line length on SE output

2.2 Correction in ICT Parameters
In transformers the tap changer types and winding
parameters were corrected at many places in database. ICT
parameters include Series Resistance, Series Reactance, Tap
Changer details (number of taps and step size), Nominal
voltage of HV and LV side, placement of Tap changer (HV
or LV side), etc.
For number of ICTs in database, it was found that the Step
size was incorrect (in magnitude or sign) and also Tap
changer was placed at both HV and LV side, which is not as
(b) Network after truncation per the field equipment. Generally, Tap changer is put on
Fig.4: Truncation of network at 400kV and above voltage level at
NLDC
HV side of transformer. Further, Resistance and Reactance
were also corrected for many ICTs as per its MVA ratings
While carrying out network truncation; the external model is and design.
ignored and a Load/Injection is modeled at the point of
truncation with the real-time MW and MVAR data of the It was found that in some stations all the ICTs are not of the
mapped to it. same MVA rating and hence possess different impedance

2 of 6
and different MW flow. In all such cases the parameters of network based on position of switching devices (i.e. Circuit
ICTs were changed accordingly. Breakers and Isolators). If the output of topology processor
2.3 Correction in Generating Unit Parameters deviates from the actual scenario at the site (due to
incorrect/unavailable status) then output of SE application
In the Network database, the Generating Units need to be would also not of much use and many-a-times would not
properly defined with number of parameters such as give a converged solution.
Maximum/Minimum MW capacity, Maximum/Minimum
MVAR capacity, MVA rating, etc. The generating unit
parameters were taken from the Annual Compendium
Booklet of POSOCO and updated in database.
All these parameters impact the State Estimator Output
considerably. An example is shown in Fig.7.

Fig.9: Impact of incorrect topology on SE output

Fig.7: Impact of incorrect Maximum/Minimum MW limits on SE First step taken towards Topology Correction is “Auto
output Replacement” of CB/Isolator positions to “CLOSED” using
3. Incorrect Mapping of switching devices between a script provided by vendor. Under this approach all
NLDC and RLDCs switching devices which have “Suspect”, “Between” or
In order to exchange data between two systems with “Garbage” status are replaced as “CLOSED” in Network
different naming convention in ICCP, it is required that data (real-time SCADA displays remain unaffected).
ICCP mapping shall be done at either of the ends. In case of
incorrect ICCP mapping the topology of the system as
changed on the site is not exactly replicated at NLDC. An
example is shown in Fig.8.

(a) Network before topology correction

(b) Network after topology correction

Fig.10: Topology Correction approach
Fig.8: Incorrect mapping of switching device at NLDC Switching devices with “Suspect” quality status leads to
4. Topology Correction point of conflict as it is not possible to make out whether the
associated line is charged or under outage/tripping.
The first step executed by almost all the State Estimator
applications worldwide is Topology Processing. Topology Status of switching devices with “Suspect” quality can be
Processor extracts the line-energised network out of the total replaced as CLOSED for State Estimator application.

3 of 6
Another approach is to keep the last good value. Different
vendors may have different approached to handle this.
Whatever approach is chosen one has to look at such device
status & validate them and get the telemetry issue fixed.
Fig.11 further elaborates this approach of correcting the
status information.

Fig.11: Topology correction Fig.13: Incorrect status of bypass switching device of Series
compensation
The status of switching devices with “Good” Quality
remains unchanged. 7. Modeling of Loads and Injections to cater HVDC
connections
5. Topology Estimation
The HVDC lines are generally modeled as Truncated
Topology estimation checks could be performed on network Loads/Injections with the real-time power flow data on the
to find ambiguity with respect to the line flows and status of convertor transformers mapped onto it. The AC filter banks
the corresponding switching devices. Coherency warnings are modeled as a set of Capacitors connected directly to the
generated could be used to tune the network. The threshold bus in order to provide the required MVARs. One of the
for analog data for generating Coherency Warning is kept as examples of +/- 500kV Mundra-Mahendragarh HVDC
50 MW. Bipole is shown in Fig.14.
The drawback of topology estimation checks in NLDC
system is that is does not consider the entire connectivity of
the equipment to the bus. It only checks the coherency with
the switching device (say Line Isolator) directly connected
with the equipment.

Fig.14: Modeling of Loads to cater HVDC flows in the network

8. Reverse Flow on parallel ICTs
In a number of stations it was observed that the MW,MVAR
flow on parallel ICTs was opposite to each other. In some
Fig.12: Topology Estimation with coherency warnings at NLDC stations it was also found that the flow on ICTs was from
LV to HV side without any source of generation at the Low
6. Status of Series Compensation
Voltage side.
Indian Power System has more than 40 nos. of series
While truncation of network the secondary of some ICTs at
compensation (FSCs and TCSCs) and the status of its by-
400/(less than 400kV) level the modeled loads at the point
pass CB/Isolator is of utmost importance in depicting
of truncation are mapped with either its primary or
whether the Series Compensation device in in service or not.
secondary value (as available). In case both side
It was found that at many stations either the status of bypass
measurements are available, then either of it can be mapped
CB/Isolator was incorrect or “suspect” leading to incorrect
to the modeled load. If the sign of value at Primary or
State Estimation at that point.
Secondary gets reversed then it may cause the modeled load
to act as Injection of power into the bus. An example is
shown in Fig.15 and Fig.16.
4 of 6
 solution the value of Load can be calculated online and
inserted into the measurement of truncated loads. The most
common way to calculate the Loads is by Bus Summation
and equating the same to Zero.

Fig.15: Impact of Reverse flow in parallel ICT on SE at NLDC in case

of truncation at that point
It can be resolved temporarily (until data sign is corrected Fig.17: Inserting calculated values on non-measured loads
from site) by changing the “FLIP” flag associated with the
Analog as “TRUE”. By doing this the incorrect value (in 10. Modeling of Equivalent Generator in special cases
sign) will be reversed before getting transferred to SE In many cases it was observed that the status of switching
application. devices associated with Units was “suspect” leading to
As per the sign convention of ICT related data in Load problem in topology Estimation whether to include the unit
Despatch Centres, different philosophy for measuring in SE process or not. In order to cater it, an equivalent
transformer data is followed in different Regions. In generator was modeled and the sum of individual units was
Northern Region, when, power flows from HV side to LV mapped to it. The equivalent generator always remains in-
side analog measurement is taken as Positive whether it is service and generates as per the “sum of all individual
measured on HV side or LV side. Whereas, in other regions units”. An example is shown in Fig.18.
the flow directions are positive when the power flows from
bus into the transformer. Hence the association of “FLIP”
flag with modelled loads shall be as per Table 1.

Fig.18: Modeling of equivalent generator

11. Tap Position of Transformers
Table 1: Association of “FLIP” flag at NLDC with measurements on
ICTs mapped on loads at points of truncation
At many sub-stations the Tap position of the ICTs is not
available with “good” quality. As the OLTC of transformers
is not frequently changed, it is manually replaced by
Conforming its position from the field level. An example is
given in Fig.19.

Fig.16: Impact of measurement mapping at modeled loads from

Primary and Secondary end (including case of sign reversal)
9. Inserting calculated values on non-measured loads
Some of the State Estimation Algorithms do not try to Fig.19: Unavailability of OLTC of ICTs
change the MW and MVAR corresponding to loads. It either 12. Re-initialising the parameters of State Estimation
uses the real-time data-telemetry assigned to it or a Pseudo
measurement taken from Load Scheduling (or any other The state estimation application has various parameters
source). If a Pseudo measurement is used then it has a high which need to be initiated after certain conditions are met
uncertainty level. In order to improve the State Estimator every time it runs. In a large power system network, it is
5 of 6
many-a-times required to re-set it after certain defined commissioned in time. So these circuits were set to
number of run-times. “manually removed” from Network.
13. Threshold of Solution cost and other parameters
Many a times it is observed that the voltage and angle
convergence is achieved while referring the analysis report,
but the Solution is declared as Invalid. Generally, it is due to
high cost of the solution exceeding the threshold limit as
defined for it. Either the threshold limit of Solution Cost is
increased temporarily or further tuning is done to handle
such problems.
Fig.22: Parallel connection of ICT in same bay
In a special case at Mandaula substation as shown in Fig.22
it was found that the spare 400/220kV ICT was connected in
parallel with ICT-1 in the same bay in order to meet the high
load in Delhi during Commonwealth Games 2010 and High
Summer Peak Load in 2015. In all such cases the parameters
of ICTs must be put accordingly.
IV. Solution Quality of SE
After continuous correction of database parameters,
connectivity model, topology errors and rectifying the bad-
data by getting it corrected from the field, the cost of SE
solution decreased significantly.
Fig.20: Threshold of Solution Cost in SE Application at NLDC
14. Unexpected connectivity changes at the site level
Due to various construction related problems (like Right of
Way i.e. RoW, Contractual issues, material management
issues, labour issues, modification in transmission planning,
etc.) the line and bay connectivity changes either
temporarily or permanently. It is many-a-times not known to
the operator and when informed requires corresponding
modeling change in SCADA and Network database.
Fig.23: Reduction in SE Solution cost
It is also observed that many lines are modelled in database
but do not get commissioned on site due to delay in V. Conclusion
construction. Such lines are set to “manually remove” from Bad data and Incorrect Network Model Database severely
Network Database. affects the output of State Estimator and it needs to be
corrected and maintained with utmost importance. As
Topology formation is an initial and critical part of SE
application but due to various reasons its telemetry is not
correctly available from site. Efforts shall be made by the
State and Central Utilities to provide the correct status of the
switching devices by rectifying the related issues at site.
Topology Estimation shall provide feature of auto-detecting
the incorrect status of switching device on the basis of
comparing the status with corresponding device MW flow
and simultaneously correcting it. After all the efforts in
tuning SE, it was found that Cost of State Estimator
Solution decreased and a better converged solution was
obtained from State Estimator Application at NLDC.
ACKNOWLEDGEMENT
Authors are grateful to the POWERGRID and POSOCO
Fig.21: Re-arrangement of connectivity at the site level management along with all the load despatchers for their
An example of temporary re-arrangement is shown in support and encouragement for continual improvement in
Fig.21. In was observed that certain sections of few 400kV quality of measurements and running of State Estimator.
lines were not complete and hence re-arrangement as shown REFERENCES
in Fig.21 above was done on temporary basis to facilitate
[1] Power generation, operation and control, Allen J. Wood and Bruce F.
early use of assets by the utilities. To compensate it, re- Wollenberg, Second edition, 1996.
modeling of lines was done in database at NLDC. [2] Power System State estimation – Theory and Implementation, Ali Abur
Moreover, two circuits of 765kV level were not and Antonio Gomez Exposito, 2004 edition.

6 of 6