State Estimation
Lamine Mili
Virginia Tech
Alexandria Research Institute
1
Control Center
: P and Q measurements
n = 2N -1 1.5 m / n 3
2
High voltage and high current
CT
0 to 5 A
I Q P V
10 k
0 to 10 V
ADC
Control Center 3
Types of Measurement Errors
Random errors - related to the class of precision of
the instrument.
500 KV Bus
T1 T2
230 KV Bus
S1 S2 S3 S4
5
Meter Values (treatments) in p.u.
1.05
S3
1.04
6 kV S4
1.03
S1,T1,T2
1.02 S2
1.01
1
0 50 100 150 200 250 300
6
Snap-shots (blocks)
Measurement Calibration
The present practice is to perform an on-site
calibration, which is rarely carried out.
j C /2 j C /2
9
Probability Distribution of Measurement Errors
f(x)
Gaussian
Actual distibution
distribution
x
0 3
10
Breakdown Point of an Estimator
The breakdown point is defined as the maximum
fraction of contamination that an estimator can handle
True value
bias mean
11
Breakdown Point of Sample Median
True value
median
median
bias median
m 1
Breakdown point of L1-norm estimator is = [
]/ m
2 12
Maximum bias curve of the sample median
Maximum Bias
0
0 0.1 0.2 0.3 0.4 0.5
Fraction of contamination
13
z = ax + b min ri
i
z
4 Vertical outlier
0
0 2 4 6 x
14
z = ax + b
min ri
z i
Bad leverage
4 point
Critical value of x
0
0 2 4 6 8 10 x
15
Leverage Points in Power Systems
These are distant points (outliers) in the space spanned
by the row vectors of the Jacobian matrix.
18
All the measurements associated with a topology
error will be seen as conforming bad data by the
state estimator. The state estimator breaks down.
19
Proposed Solution
Develop a preprocessing method that does not
assume that the topology as given.
In this model, the state variables are the power
flows of all the branches, be they energized or not.
20
xPi
Pkl x pi
Vk Vl 1 pu
kl X kl xPi
21
Topology estimator
22
Parameter estimator
Take advantage of the fact that the state remains
nearly unchanged over a certain period of time,
typically during the late night off-peak period.
Measurement placement