EE 369

POWER SYSTEM ANALYSIS

Lecture 13
Newton-Raphson Power Flow
Tom Overbye and Ross Baldick

1
 Announcements
• Homework 10 is: 3.49, 3.55, 3.57, 6.2, 6.9,
6.13, 6.14, 6.18, 6.19, 6.20; due 11/17. (Use
infinity norm and epsilon = 0.01 for any
problems where norm or stopping criterion
not specified.)
• Homework 11 is 6.24, 6.26, 6.28, 6.30, 6.38,
6.42, 6.43, 6.46, 6.49, 6.50; due Tuesday
11/22. Note that HW is due on Tuesday
because Thanksgiving is on Thursday.
•2
 Dishonest Newton-Raphson
Since most of the time in the Newton-Raphson
iteration is spent dealing with the Jacobian, one
way to speed up the iterations is to only calculate
(and factorize) the Jacobian occasionally:
– known as the “Dishonest” Newton-Raphson or
Shamanskii method,
– an extreme example is to only calculate the Jacobian
for the first iteration, which is called the “completely
dishonest Newton-Raphson” or chord method.
Honest: x ( v 1)  x ( v ) - J (x ( v ) )-1 f (x ( v ) )
Dishonest: x ( v 1)  x ( v ) - J (x (0) )-1 f (x ( v ) )
Stopping criterion f ( x ( v ) )   used in both cases. 3
 Dishonest Newton-Raphson
Example
Use the Dishonest Newton-Raphson (chord method)
to solve f ( x )  0, where:
f ( x)  x2 - 2
1
x (v)
   df (0) 
(x ) f ( x(v) )
 dx 

x ( v )    1  (v) 2
(( x ) - 2)
 2 x 
(0)

(v)  1 
x ( v 1)  x  (( x (v) 2
) - 2)
 2 x (0) 
4
Dishonest N-R Example, cont’d
x ( v 1)
 x (v)  1  (v) 2
  (0)  (( x ) - 2)
2x 
Guess x (0)  1. Iteratively solving we get
 x ( v ) (honest) x ( v ) (dishonest)
0 1 1 We pay a price
1 1.5 1.5 in increased
2 1.41667 1.375 iterations, but
with decreased
3 1.41422 1.429 computation
4 1.41422 1.408 per iteration
5
 Two Bus Dishonest ROC
Region of convergence for different initial
guesses for the 2 bus case using the Dishonest N-R
Red region
converges
to the high
voltage
solution,
while the
yellow region
converges
to the low
voltage
solution

6
Honest N-R Region of Convergence

Maximum
of 15
iterations

7
 Decoupled Power Flow
The “completely” Dishonest Newton-
Raphson (chord), where we only
calculate the Jacobian once, is not
usually used for power flow analysis.
However several approximations of the
Jacobian matrix are used that result in a
similar approximation.
One common method is the decoupled
power flow. In this approach
approximations are used to decouple the
real and reactive power equations. 8
Coupled Newton-Raphson Update
Standard form of the Newton-Raphson update:
 P ( v ) P ( v ) 
   θ( v )  
 θ V  P( x ( v ) ) 
       f ( x (v)
)
 Q ( v ) Q (v) 
  V
(v)
  Q ( x (v)
) 
 
 θ  V 
 P2 ( x ( v ) )  PD 2  PG 2 
 
where P( x )  
(v)
.
 P (x( v ) )  P  P 
 n Dn Gn 
Note that changes in angle and voltage magnitude
both affect (couple to) real and reactive power. 9
 Decoupling Approximation
P ( v ) Q ( v )
Usually the off-diagonal matrices, and
V θ
are small. Therefore we approximate them as zero:
 P ( v ) 
 0 
 θ (v) 
 P ( x ( v ) ) 
θ
        f ( x (v)
)
 Q    V 
(v) ( v )

 Q( x ) 
( v )
 0  
 V 
Then the update can be decoupled into two separate updates:
( v ) 1 ( v ) 1
 P  (v)  Q 
θ (v)
    P ( x (v)
), V     Q ( x (v)
).
 θ   V  10
 Off-diagonal Jacobian Terms
So, angle and real power are coupled closely, and
voltage magnitude and reactive power are coupled closely.
Justification for Jacobian approximations:
1. Usually r x, therefore Gij Bij
2. Usually ij is small so sin ij  0
Therefore
Pi
 Vi  Gij cosij  Bij sin ij   0
 Vj
Qi
  Vi V j  Gij cosij  Bij sin ij   0
θ j 11
Decoupled N-R Region of
Convergence

12
 Fast Decoupled Power Flow
By further approximating the Jacobian we obtain
a typically reasonable approximation that is
independent of the voltage magnitudes/angles.
This means the Jacobian need only be built and
factorized once.
This approach is known as the fast decoupled
power flow (FDPF)
FDPF uses the same mismatch equations as
standard power flow so it should have same
solution if it converges
The FDPF is widely used, particularly when we
only need an approximate solution. 13
 FDPF Approximations
The FDPF makes the following approximations:
1. Gij  0
2. Vi  1 (for some occurrences),
3. sin ij  0 cos ij  1
Then: θ( v )  B 1diag{| V |( v ) }1 P( x ( v ) ),
 B 1diag{| V |( v ) }1 Q( x ( v ) )
(v)
V
Where B is just the imaginary part of the Ybus  G bus  jB bus ,
except the slack bus row/column are omitted. That is,
B is B bus , but with the slack bus row and column deleted.
Sometimes approximate diag{| V |( v ) } by identity. 14
 FDPF Three Bus Example
Use the FDPF to solve the following three bus system
Line Z = j0.07

One Two

200 MW
100 MVR
Line Z = j0.05 Line Z = j0.1

Three 1.000 pu

200 MW
100 MVR  34.3 14.3 20 
Ybus  j  14.3 24.3 10 
 
 20 10 30 
15
FDPF Three Bus Example, cont’d
 34.3 14.3 20 
   24.3 10 
Ybus  j 14.3 24.3 10  B   
   10 30 
 20 10 30 
 0.0477 0.0159 
B 1   
 0.0159 0.0389 
Iteratively solve, starting with an initial voltage guess
(0)
 2 
(0)
0  V 2  1
    V   1
 3 0   3 
 2 
(1)
0   0.0477 0.0159   2   0.1272 
         
 3   
0 0.0159 0.0389   
2 0.1091
16
FDPF Three Bus Example, cont’d
(1)
V 2  1  0.0477 0.0159  1  0.9364
 V   1   0.0159 0.0389  1   0.9455
 3       
Pi ( x ) n P P
  Vk (Gik cosik  Bik sin ik )  Di Gi
Vi k 1 Vi

 2
(2)
 0.1272   0.0477 0.0159   0.151  0.1361
    0.1091   0.0159 0.0389  0.107    0.1156
 3       
(2)
V 2  0.924 
V    
 3  0.936
 0.1384  0.9224 
Actual solution: θ    V 
 0.1171   0.9338 
17
FDPF Region of Convergence

18
 “DC” Power Flow
The “DC” power flow makes the most severe
approximations:
– completely ignore reactive power, assume all the
voltages are always 1.0 per unit, ignore line
conductance
This makes the power flow a linear set of
equations, which can be solved directly:
θ   B 1 P
where B is the imaginary part of the bus
admittance matrix with the row and column
corresponding to the slack bus deleted, and,
similarly, Θ and P omit the slack bus. 19
DC Power Flow Example

20
 DC Power Flow 5 Bus Example

One Five Four Three

A A

360 MW 520 MW
MVA MVA
A

0 Mvar
MVA

slack
0 Mvar

1.000 pu 1.000 pu A A
1.000 pu 80 MW
0.000 Deg -4.125 Deg MVA MVA
-1.997 Deg 0 Mvar
1.000 pu
0.524 Deg

1.000 pu Two
-18.695 Deg

800 MW
0 Mvar

Notice with the dc power flow all of the voltage magnitudes are
1 per unit.
21
 Power System Control
A major problem with power system operation is
the limited capacity of the transmission system
– lines/transformers have limits (usually thermal)
– no direct way of controlling flow down a
transmission line (e.g., there are no low cost valves
to close to limit flow, except “on” and “off”)
– open transmission system access associated with
industry restructuring is stressing the system in new
ways
We need to indirectly control transmission line
flow by changing the generator outputs. 22
Indirect Transmission Line Control
What we would like to determine is how a change in
generation at bus k affects the power flow on a line
from bus i to bus j.
The assumption is
that the change in
generation at bus k
is matched by an
opposite change at
the slack bus.

23
 Power Flow Simulation - Before
•One way to determine the impact of a generator
change is to compare a before/after power flow.
•For example below is a three bus case with an
overload.
131.9 MW

124%

One Two

200.0 MW 200 MW
68.1 MW 68.1 MW
71.0 MVR 100 MVR

Z for all lines = j0.1

Three 1.000 pu

0 MW
64 MVR 24
 Power Flow Simulation - After
•Increasing the generation at bus 3 by 95 MW
(and hence decreasing generation at the slack
bus 1 by a corresponding amount), results in a
31.3 MW drop in the MW flow on the line from
bus 1 to 2. 101.6 MW

100%

One Two

105.0 MW 200 MW
3.4 MW 98.4 MW
64.3 MVR 100 MVR

92%
Z for all lines = j0.1
Limit for all lines = 150 MVA
1.000 pu
Three
95 MW
64 MVR
25
Analytic Calculation of Sensitivities
Calculating control sensitivities by repeated
power flow solutions is tedious and would
require many power flow solutions.
An alternative approach is to analytically
calculate these values
The power flow from bus i to bus j is
Vi V j i   j
Pij  sin( i   j ) 
X ij X ij
 i   j  ij
So Pij  We just need to get
X ij PGk26
 Analytic Sensitivities
1
From the fast decoupled power flow we know: θ  B P( x ).
Sign convention in definition of P( x ) is that entry in P( x )
is negative if change in net injection (generation) is positive.
So to get the change in θ due to a change of generation at
bus k , just set P( x ) equal to all zeros except a minus one
at position k :
0
 
 
P   1  For 1MW increase in generation at bus k
0
 
  27
 Three Bus Sensitivity Example
For the previous three bus case with Zline  j 0.1
 20 10 10 
   20 10 
Ybus  j 10 20 10  B   
   10  20 
 10 10 20
Hence for a change of generation at bus 3
1
  2   20 10   0   0.0333
        
 3  10  20 
   1 0.0667 
0.0667  0
Changes in line flows are: P3 to 1   0.667 pu
0.1
P3 to 2  0.333 pu P 2 to 1  0.333 pu 28