Transcript State Estimator Performance

State Estimator performance
1
State of Technology
• Many metrics have been proposed for measuring
State Estimator performance..
PI = min Σ(Z – hi(x))^2
m
Mmeasx^2= Pr(x2m-n > Σ(Zi –hi(x))^ 2
^2
i=1
^2
• But these metrics are generally designed to test the
effectiveness of the state estimator algorithm in
using available data…That is not our purpose
2
What is our purpose
• We are implementing a new market design. We
are not interested in the technical merits of one
State Estimator over another; but as good
capitalists we are interested in getting the price
right.
• So – What makes the price right in a nodal
market.
3
The price is right when predicted congestion
matches reality (observed post contingency
overloads)
37 MW
15Mw
37 MW
(15)
15Mw
(15)
~
23Mw
FAULT
8Mw
25Mw
(20)
17 MW
8Mw
40Mw
(28)
0 MW
0 MW
25Mw
(17)
17Mw
4
(40)
Why State Estimator?
• The original purpose of state estimation was to
allow operators to maintain reliable operation of
the transmission system.
• State Estimators facilitate this ability by providing
what NERC calls “situational awareness”.
• By assessing the “state” of the transmission
system, State Estimator allows us to create a
power flow model of the transmission system
which can be used to explore “what if” scenarios.
What will happen if this element fails…
5
Why State Estimator?
• Nodal markets are designed to efficiently operate
a transmission/generation system in a way that
economically considers transmission congestion
issues in generation dispatch decisions; and
reflects those constraints in prices.
• Including the “what if” scenarios considered for
reliability is central to this design.
6
Do we have to get everything
perfect to have it “right”?
• For efficiency we want to expend our limited
resources on improvements only where it is most
effective.
• How do we target our improvements where it is
most cost effective?
• Answer – Identify locations where congestion has
been observed historically and target our
improvements to these locations
7
How have other markets
addressed this issue?
• We have not identified SE performance requirements
for PJM, New England or New York
• Mid West ISO Targeted State Estimator performance
in multiple ways:
–
–
–
–
–
–
–
–
30 day test with 97% convergence
ICCP & RTU monthly availability 99.5% of time end to end.
50 most limiting elements SE-telemetry residual < 5% or 50 MW
50 units SE-telemetry < 5% or 50 MW
50 flowgates SE-telemetry < 1%
50 critical buses SE voltage within 2% of metered
10 tielines SE-telemetry < 5% or 50 MW
All other branches >100kV SE-telemetry within 10%
8
Proposed SE Requirement 1:
• State Estimator converges 97% of runs during a 30
day test period.
– Objective is met if SE converges 97% of runs during a designated
30 day test period.
9
Proposed SE Requirement 2:
• On transmission elements identified as causing 80%
of congestion cost in 2006, ERCOT will reduce the
MW telemetry – MW SE residual to < 3% of the
associated line emergency rating (based upon 3%
accuracy requirement)
– Objective met if MW residual of specific identified elements is <
3% of the associated element rating on over 95% of samples
measured in a one month trial. After commercial operation; to be
measured on all State estimator samples of each month and
posted.
10
Proposed SE Requirement 3:
• On transmission elements identified as causing 80%
of congestion cost in 2006, ERCOT will confirm that
the MW base case element flows used in contingency
analysis, and the MW state estimator line flows used
in creating that power flow, differ less than 3% of the
equipment emergency rating
– Objective met if difference of all identified elements < 3% of
equipment rating over 95% of samples measured during a 1
month trial. To be measured on all State estimator samples of
each month and posted.
11
Proposed SE Requirement 4:
• On 20 most critical buses designated by ERCOT and
approved by TAC each October; Telemeter bus
voltage minus state estimator voltage within 2% or the
accuracy of the voltage measurement involved.
– Objective met if difference of all identified voltages < equipment
accuracy or 2% of nominal voltage over 95% of samples
measured during a 1 month trial. To be measured on all State
estimator samples of each month and posted.
12
Proposed SE Requirement 5:
• On all transmission elements > 100KV; the difference
between state estimator MW solution and the scada
measurement will be less than 10 MW or 10% of the
associated emergency rating (whichever is greater)
on 99.5% of all equipment. All equipment failing this
test will be reported to the associated TSP for repair
within 10 days of detection.
– Objective met if during 30 day trial period prior
to market operation 99.5% of flow
measurements
13
To meet these objectives
• ERCOT will work with TSP to resolve problem in
accordance with ERCOT Guides prior to directing
additional equipment.
• In event of failure to meet the standard: ERCOT may direct
additional telemetry be installed on lines contributing most
to 80% of congestion costs for 2006 to meet the 3% of
equipments emergency rating. (related to accuracy
requirement specified in the protocols. 6.5.5.2 (7))
• ERCOT will enforce the requirements of 3.10.7.4.2 (5) by
alarming any sum of flow around a bus > 5% or 5MW, and
requesting that the applicable TSP or QSE correct the
failure.
14
Discussion
Issues
•
•
•
•
•
Problem – How does TSP know if it is a TSP problem or an ERCOT
modeling problem.
It sounds like you want the residual tolerance to be a function of
the line rating – yes
I would rather see TSP’s work with ERCOT to reduce the ERROR
rather than automatically install new telemetry – I agree
Throughout section 3 10 mw seems to be a magic number. There
seems to be inconsistency with the 5 MW/5% discussion
AEN most cases we have investigated we have observed that the
input data (scada) is incorrect; perhaps we should be sure the
system is not imbalanced. On one hand we might need to estimate
parameters; but there are so many other problems in the system.
When we have identified problems; most of the time it is due to
imbalances – vector of MW/MVAR do not match MVA. Sometimes
this is due to current imbalances between 3 phases.
15
Discussion
Issues
• What should ercot do when the 3% criteria is
violated?
– Centerpoint does not often see the gap between SE and Power flow which
ERCOT says it is experiencing. Suggests that this may be due to modeling
problems instead of the telemetry accuracy.
• ERCOT shall contact TSP/QSE to verify the data;
shall arrange for the data provider to view the SE
and telemetry.
16
Agreed Upon
1) SE 97% convergence of runs during a 30 day test period.
2) On transmission elements identified as causing 80% of congestion cost in 2006 or top
30 lines costing congestion whichever is fewer, ERCOT will reduce the MW telemetry
– MW SE residual to < 3% of the associated line emergency rating.
3) On transmission elements identified as causing 80% of congestion cost in 2006 or top
30 lines costing congestion whatever is fewer, ERCOT will confirm that the MW base
case element flows used in contingency analysis, and the MW state estimator line
flows used in creating that power flow, differ less than 3% of the equipment
emergency rating. (Note – NDSWG is concerned that the current performance of all
lines according to ERCOT staff; is 32% of lines are over 4.6% different on this metric.
ERCOT has not yet been able to supply NDSWG with the subset of lines which would
be measured under this metric. We are assuming ERCOT with TSPs can improve
performance to meet this metric. This is a risk.)
4) On 20 most critical buses designated by ERCOT and approved by TAC each October;
Telemeter bus voltage minus state estimator voltage within 2% or the accuracy of the
voltage measurement involved.
5) On all transmission elements > 100KV; the difference between state estimator MW
solution and the scada measurement will be less than 10 MW or 10% of the
associated emergency rating (whichever is greater) on 99.5% of all equipment. All
equipment failing this test will be reported to the associated TSP for repair within 10
days of detection.
6) Each TSP shall provide telemetered measurements on modeled Transmission
Elements to ensure State Estimator observability of any monitored voltage and power
flow between their associated transmission breakers to the extent such can be shown
to be needed in achieving items 1-5 above. On monitored non-Load Electrical Buses,
each TSP shall install, at the direction of ERCOT, sufficient telemetry such that there
is an “N-1 Redundancy.” An N-1 Redundancy exists if the Electrical Bus remains
observable on the loss of any single measurement pair (kW, kVar) excluding station
RTU communication path failures. In making the determination to request additional
telemetry, ERCOT shall consider the economic implications of inaccurate
representation of Load Models in LMP results
versus the cost to remedy.
17
Assignments
• Assignments
– ERCOT to provide list of lines which are expected to qualify as in the 80%
and compare to the single case comparison they have.
– ERCOT propose 20 buses for voltage. Corpus, Houston, DFW, Valley,
Laredo, Bryan?, Killeen?, 2 buses each area…
– ERCOT to propose observability standard and definition.
– ERCOT to send draft SE and telemetry standard to David.
– Participants to comment on Section 8 re-write and send to David.
18