SALAM

What is LFNewton()?

Load Flow using Newton Rophson Method

Before anything I suggest you to study the LFGauss.pdf file which is included in
LFGauss().zip package. By the way this package is completely independent.
As we will see through an example, learning how the program works is pretty easy. The same
as LFGausse the formulas used in it are based on Power System Analysis by Hadi Saadat
(more info.)
Actually the program uses a lot of RAM space if the network has more that 6 buses but its
not something to worry in TI-89 Titanium, you have much more RAM than needed, so I didnt
compress the Ybus and vpqkxmvp matrices. (The second one is used to store information
about voltages, active and reactive power and other characteristics of buses.)

How to install:
Just send the shin_adm and LFNewton files to you TI device. (Any folder but I use PSA
for power system analyzing programs and functions.)
The shin_adm() is a program that generates YBUS matrix. This matrix will be stored in
variable xxx and this is the only variable that is used by the program, other variables are
local.
Another thing to mention is that both programs can be archived after running once, but the
variable xxx must be unlocked so shin_adm can store YBUS in it.

OK, lets see how it works:

Imagine this system:
Before anything I suggest you to change all impedance values to admittance:
y12=10-20j
y13=10-30j
y23=16-32j
If you dont do this here, you will have to enter them in the program like this:
y12= (0.02+0.04j) ^-1
1. Run the program LFNewton() - (no parameters are needed.)

(& Press Enter)

2. The program is asking you about Ybus, if the question has given this matrix to you
just store it in this variable and then when you reach hear enter 1, otherwise you
should enter any number except 1 so the program automatically calls the other
program shin_adm(n) and after a few steps it stores the Ybus in xxx. (n is the
number of buses in network.)
We assume that Ybus is not stored in xxx so for example I enter 0:

3. Enter number of buses. (3 in the above network)

4. Entering the line data.

y [ i , j ] = the admittance between i and j buses. (Note that its different from Yij
in Ybus matrix.)
The parallel admittance (if it exists in your line model) is not used here. You
should just enter the amount of connected admittance to both i and j buses.
y [ i , i ] = the admittance between bus i and the earth (ground). If the line
doesnt have parallel admittance (that is caused by capacitor between line and
the earth) this amount is zero.

So follow the pictures:

5. Up to now Ybus is created and stored in xxx.
(You can generate this matrix in the same way by running shin_adm(n) program.)

As we saw the shin_adm(n) is a very usefull program to generate Ybus of a small

network but if the network has more than 5 buses then this method of inputting
data is not efficient, specially in real world problems. If you want to know why, read
the blue text below:

Another method of inputting line data:

Another way is to input line data like this:
[1,2,1+10j;1,3,4-5j;;3,2;6-7j]
And after inputting this matrix the program will understand that for example the
admittance between bus 1 and bus 2 is 1+10j, and the same for bus 1 and 3, (4-5j),
and all the other combinations which are not included are zero.
If you test both methods you will find out that the one shown in forth step is faster
for a small network. (Just try to input that matrix), but when we face a real world
problem, its totally different. Because many of buses are not connected to each
other so you have to input zero many times; in this situation inputting line data in the
form shown here is much faster. But our program is written to solve a problem in a
test, and these kind of problems wont have more than 5 buses!
Compressing Ybus:
In real world problems we have lots of buses, for example, imagine a network with
30 buses, but every bus is only connected to a few other buses (usually something
between 2 to 4 buses).
Now most of the element of Ybus matrix become zero, in the above example Ybus is
a 30*30 matrix, but each row has only 2 to 4 non zero elements, because the
admittance between bus i and most of the other buses is zero (there is no line
between them). In the past, computers didnt have enough memory to store this BIG
matrix, so programmers suggested some ways to compress Ybus (its not
complicated at all, actually all zip programs do the same work, you can even write a
program yourself to compress this matrix), but nowadays that computers have lots of
memory it not needed to compress this matrix at all. (If you want to learn more
about compressing this matrix I suggest you to study Electric energy systems theory
by O.I. Elgerd)

Now is the time to enter bus data values:

(Im sure you know what is a PQ or a PV bus but just as a reminder, Slack bus, is a
bus which is going to compensate the difference between load and generation. These
kinds of power stations should have special characteristics that Im not going to
explain here. A PV bus is one which has a generator except the slack bus. PQ buses
are the rest, they only have loads. Be careful that a capacitor bank changes a bus
into generating bus or PV, because the capacitors generate reactive power.)

For every bus we assign a number, when using this program use number one for
slack bus. (Its not a limit for program because the numbers are optional and in the
other hand I can say that this is a standard, even if the question has given a different
number for this bus just change the number with the number of a bus that is one.)
On basis of the questions picture above, the PQ bus number is 2 and PV is 3.

Another important point when using this or any other method of load flow is that
everything should be per-unite. (The base value of power in this system is 100MVA)

As you see the program automatically finds the number of PV buses, so you just
need to enter bus number.
Most of generating buses have a reactive power generation limit, if there is also a
load on the PV bus use Qmin=Qm-Qd in which Qm is the minimum generated
reactive power by the generator and Qd is the demanded reactive power on this bus.
If there is no limit to minimum reactive power for the generator, use -.

This is the absolute of Ybus matrix. If your system has more than 3 buses it probably
wont fit on the screan, but dont worry, once you have finished you can generate
this matrix by typing abs(xxx) at home screan. The same can done about angle of it:
(angle(xxx))

(The program is showing the formulas which are going to be used. |Yij| and ij are
the element of the matrixes shown in previous screens.)
If you have downloaded this program you probably are familiar with Newton-
Rophson method in load flow so you should know what is the usage of Qs and Ps,
if you dont refer to the Saadats book.
As you can see bus number (i) and iteration counter (k) are shown in the matrix
below the screen.

These are the formulas to calculate the elements of Jacobean matrix. In some
references this matrix is shown like this:
[H, N; J, L]
But here we show them like this:
[J1, J3; J2, J4]
(The same as Hadi Saadat)
(Finished Calculating J1, the same process for J2, J3 and J4)

At the top of the screen you see the answer of sigma when calculating J2 [1, 1].

In this example J2 had only two elements.

Calculating Jacobean matrix is finished.

(The Jacobean inverse is shown in this screen; if your Jacobean matrix doesnt fit to
screen then if you want to generate this matrix you should do it manually in the main
menu on basis of the Jacobean elements you have, however in usual problems (in
standard tests I mean) the question is made in a way that usually this matrix wont
exceed 3*3, for example if you have another PV in this question, then the row and
column of that bus in Jacobean matrix will be omitted so your Jacobean would have
only one element!
Do you know why?
Because in PV buses the voltage is constant (its specified in the question) and in the
other hand reactive power is constant too so we wont have V and Q.
If you still dont get what I mean or why Q=0 just email me.)
OK! First iteration finished.
A little boring huh? Yes, Newton-Rophson method has long steps but only a few
steps are needed to reach the final answer.
For example if you have solved this problem using GS method then you know
reaching the answer with an error of less than 5*10^-5, need at least seven steps
but here we will see that only after 3 steps we will reach the final answer.
But there are other advantages that make NR much more reliable. For instance in a
few system the iteration may not become converge when using GS method but the
probability for such situation is very less in NR method. (Besides the method you use
being convergent or not is very relevant to your first guess of voltages; as a standard
in programs we assume that all PQ voltages are 1p.u. because in a real system and
its usual work (I mean except short circuit situations) it is really something near
1p.u. .)

If you want to calculate more Vis with more repeats enter 1, any other number will
lead you to next loop.


Only one more step!

We wanted to know how much power should the slack bus produce. In some kind of
problems the power produced by other PV buses is questioned too, so the program is
designed in a way that can give you the power of any bus!

The rest of program just gives a few useful information about the system (Iij, Sij and
Sloss,ij) in which Sloss,ij is the Power loss in line i to j.

After you finish you can see the Ybus by typing xxx:

Thats all.
The only limit I found when using this program was that its not programmed to
calculate the effect of tab-changing of transformers. (Of coarse there is a big
difference between a real system and what we just explained, but as a program to
help you solve your problems I think its very useful.)

Other related programs:

I have also written 3 other programs which use Newton-Rophson, Decoupled and
Fast Decoupled methods to solve load flow problem. (The DC load flow is too simple
and Im not going to program it.)
Here is a list of programs Ive written:
Load flow using Gauss-Seidel method.
Load flow using Decoupled method.
Load flow using fast decoupled method.
SCTM (Symmetrical components transformation matrix created by doctor
C.L.Fortescue.) & SCTMI (Inverse of SCTM)
puBasech( (a function to calculate per unit values after change of base
value.)
linegc() (calculates line general coefficients for short and medium length
lines.)
llinegc() the same as linegc() but for long lines (uses hyperbolic functions.)
GMD, GMR, GMRb (Geometric mean distance and radius of lines, these
programs are slow I will be very thankful if you can send me any suggestions
to improve these programs.)
OCT() and SCT(). (Open circuit test and short circuit test for a DC machine.)
IEEE (finds IEEE model of Induction machine.)
W2D and D2W (delta<->why )

Sorry for my bad English, if there is anything that I should add or any dictation or grammar
errors just email me. Thanks. ;D
Any suggestions, any errors, anything, just contact me:
mailto:[email protected]

Ali Dehghan Banadaki

(A university student in Islamic Azad University of Gonabad.)
Thursday, February 07, 2008