I'm interested in some matrices associated with the pnnl feeders, namely:
A) The impedance matrix
B) The power flow jacobian matrix (delta P, delta Q wrt delta V angle, delta V mag)
Questions:
1) Is there a way to extract these matrices?
2) Perhaps (B) could be extracted by providing an external LU interface - is this so? My C skills are rusty - is there some external LU code out there that I could pull from?
Thanks!
tw
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
This may be a little difficult. First, depending on the powerflow solution method you are using, there may not be a classical impedance matrix. The Forward Backward Sweep (FBS) method is an iterative ladder technique, which doesn't actually formulate an impedance matrix (since it assumes the system is radial). I'll refer you to William Kersting's book for more information:
In the second powerflow method, Newton-Raphson Current Injection Method (NCIM by trade, NR in GridLAB-D) we use a more classical Jacobian formulation. However, it uses delta I-real and delta I-imag, rather than Ps and Qs, and delta V-real and delta V-imag rather than polar. This method was developed by some really smart folks down in Brazil (Araujo, Carneiro, Costa, Garcia, Penido, Pereira, various others). I'll recommend their IEEE Transaction papers for a good resource.
Or, you can look at our very short descriptions of each here:
All that said, you can pull out the admittance matrix formed by the NR method, but you will have to compile the GLD base code yourself and activate a debug flag. In solver_nr.cpp, there is a debug flag called NR_MATRIX_OUT. If activated, this will dump a sparse-formatted, text version of the admittance matrix. It's a little hard to read, because the order of the matrix doesn't relate to the topology of the system, but rather in a way which speeds up the solving process. I believe there are also flags in SuperLU that will allow you to dump the matrix, but again, the order will be a little odd.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all,
I'm interested in some matrices associated with the pnnl feeders, namely:
A) The impedance matrix
B) The power flow jacobian matrix (delta P, delta Q wrt delta V angle, delta V mag)
Questions:
1) Is there a way to extract these matrices?
2) Perhaps (B) could be extracted by providing an external LU interface - is this so? My C skills are rusty - is there some external LU code out there that I could pull from?
Thanks!
tw
Twainwright,
This may be a little difficult. First, depending on the powerflow solution method you are using, there may not be a classical impedance matrix. The Forward Backward Sweep (FBS) method is an iterative ladder technique, which doesn't actually formulate an impedance matrix (since it assumes the system is radial). I'll refer you to William Kersting's book for more information:
http://books.google.com/books/about/Distribution_System_Modeling_and_Analysi.html?id=Y4PvKaNsWBgC
In the second powerflow method, Newton-Raphson Current Injection Method (NCIM by trade, NR in GridLAB-D) we use a more classical Jacobian formulation. However, it uses delta I-real and delta I-imag, rather than Ps and Qs, and delta V-real and delta V-imag rather than polar. This method was developed by some really smart folks down in Brazil (Araujo, Carneiro, Costa, Garcia, Penido, Pereira, various others). I'll recommend their IEEE Transaction papers for a good resource.
Or, you can look at our very short descriptions of each here:
https://sourceforge.net/apps/mediawiki/gridlab-d/index.php?title=Power_Flow_Guide
All that said, you can pull out the admittance matrix formed by the NR method, but you will have to compile the GLD base code yourself and activate a debug flag. In solver_nr.cpp, there is a debug flag called NR_MATRIX_OUT. If activated, this will dump a sparse-formatted, text version of the admittance matrix. It's a little hard to read, because the order of the matrix doesn't relate to the topology of the system, but rather in a way which speeds up the solving process. I believe there are also flags in SuperLU that will allow you to dump the matrix, but again, the order will be a little odd.