Test Case Archive of Optimal Power Flow (OPF) Problems with Local Optima

How to use this archive

Whats in it

• case2
2 bus network with one generator and 1 transmission line. Two local solutions exist within [0.95,1.05] voltage bounds.


• case3
3 bus network with 1 generator, 2 transmission lines. Two local solutions exist within [0.95,1.05] voltage bounds.


• case5
5 bus network with 2 generators, 6 transmission lines. Two local solutions exist within [0.95,1.05] voltage bounds.


• case9
9 bus network with 3 generators, 9 transmission lines. Four local solutions exist with default voltage bounds(±10%).


• case30loop
Loop network of 30 buses, 15 generaots and 30 transmission lines. Two local solutions occur within [0.95,1.05] voltage bounds.


• case39mod1
Modified from 39 bus New-England test network. Real and reactive power demands are halved. 2 local solutions exist within [0.95,1.05] voltage bounds.


• case39mod2
Modified from 39 bus New-England test network. Real and reactive power demands are halved, and only linear cost coefficients are considered. 16 local solutions exist within [0.95,1.05] voltage bounds.


• case118mod
Modified from 118 bus IEEE test network. Real and reactive generation bounds are scaled by 7. 3 local solutions exist with default voltage bounds(±6%).


• case300mod
Modified from 300 bus IEEE test network. Real and reactive generation bounds, and line limit scaled by 4. 2 local solutions exist within [0.90,1.10] voltage bounds.

Running test cases

Download the test case in your MATPOWER[1] folder. PF and OPF can be run by the commands runpf('test'), runopf('test') respectively. By default MATPOWER use PDMIP or MIPS to solve OPF problem. Note that these solvers does not take user defined intial points. If you want to reproduce the local solutions reported here use 'fmincon' as a solver and give a staring point close to the reported local solutions, or use other solvers which can take user defined initial points. We have checked the optimality of the local solutions using KNITRO, IPOPT and SNOPT with our AMPL ACOPF model.

Using data for local solutions of test cases

The data of local solutions is in MAT format. The data can be loaded from MAT-file into workspace by the command load('filename') in command window of MATLAB. Data is contained in a structure 'ACsol' of dimensions '1xnL', where nL is the number of local solutions. ACsol contains the complete solution of AC-Optimal power flow problem.

If you are using any package/modelling language other than MATLAB then you can write the relevant data file using CSV (or TXT,MAT) file format. A script to convert MAT<->CSV and MAT<->TXT file format can be be found on main page of this archive.