Test Cases and Parsers

Directory

ANDES comes with several test cases in the andes/cases/ folder. Currently, the Kundur's 2-area system, IEEE 14-bus system, NPCC 140-bus system, and the WECC 179-bus system has been verified against DSATools TSAT.

The test case library will continue to build as more models get implemented.

A tree view of the test directory is as follows.

cases/
├── 5bus/
│   └── pjm5bus.xlsx
├── GBnetwork/
│   ├── GBnetwork.m
│   ├── GBnetwork.xlsx
│   └── README.md
├── ieee14/
│   ├── ieee14.dyr
│   └── ieee14.raw
├── kundur/
│   ├── kundur.raw
│   ├── kundur_aw.xlsx
│   ├── kundur_coi.xlsx
│   ├── kundur_coi_empty.xlsx
│   ├── kundur_esdc2a.xlsx
│   ├── kundur_esst3a.xlsx
│   ├── kundur_exdc2_zero_tb.xlsx
│   ├── kundur_exst1.xlsx
│   ├── kundur_freq.xlsx
│   ├── kundur_full.dyr
│   ├── kundur_full.xlsx
│   ├── kundur_gentrip.xlsx
│   ├── kundur_ieeeg1.xlsx
│   ├── kundur_ieeest.xlsx
│   ├── kundur_sexs.xlsx
│   └── kundur_st2cut.xlsx
├── matpower/
│   ├── case118.m
│   ├── case14.m
│   ├── case300.m
│   └── case5.m
├── nordic44/
│   ├── N44_BC.dyr
│   ├── N44_BC.raw
│   └── README.md
├── npcc/
│   ├── npcc.raw
│   └── npcc_full.dyr
├── wecc/
│   ├── wecc.raw
│   ├── wecc.xlsx
│   ├── wecc_full.dyr
│   ├── wecc_gencls.dyr
└── wscc9/
    ├── wscc9.raw
    └── wscc9.xlsx

MATPOWER Cases

MATPOWER cases has been tested in ANDES for power flow calculation. All following cases are calculated with the provided initial values using the full Newton-Raphson iterative approach.

Note:

The 70K and the USA synthetic systems have difficulties to converge using the provided initial values. One can solve the case in MATPOWER and save it as a new case for ANDES. For example, the SyntheticUSA case can be converted in MATLAB with

mpc = runpf(case_SyntheticUSA)
savecase('USA.m', mpc)

And then solve it with ANDES from command line:

andes run USA.m

The output should look like

-> Power flow calculation
Sparse solver: KLU
Solution method: NR method
Power flow initialized.
0: \|F(x)\| = 140.5782767
1: \|F(x)\| = 29.61673314
2: \|F(x)\| = 4.161452394
3: \|F(x)\| = 0.2337870537
4: \|F(x)\| = 0.001149488448
5: \|F(x)\| = 3.646516689e-08
Converged in 6 iterations in 1.6082 seconds.

Performance

The numerical library used for sparse matrix factorization is KLU. In addition, Jacobians are updated in place kvxopt.spmatrix.ipadd. Computations are performed on WSL2 Ubunbu 20.04 with AMD Ryzen 9 5950X, 64 GB 3200 MHz DDR4, running ANDES 1.5.3, KVXOPT 1.2.7.1, NumPy 1.20.3, and numba 0.54.1. NumPy and KVXOPT use OpenBLAS 0.3.18. Numba is enabled, and the generated code are precompiled. Network connectivity checking is turned off. Time to read numba cache (~0.3s) is not counted.

The computation time may vary depending on operating system and hardware. All the cases are original in MATPOWER 7.0. Cases not listed below will not solve with ANDES 1.5.3.

File Name	Converged?	# of Iterations	ANDES Time [s]
case1354pegase.m	1	4	0.034
case13659pegase.m	1	5	0.276
case14.m	1	2	0.009
case145.m	1	3	0.014
case15nbr.m	1	17	0.024
case17me.m	1	3	0.010
case18.m	1	3	0.011
case1888rte.m	1	2	0.025
case18nbr.m	1	18	0.026
case1951rte.m	1	3	0.031
case2383wp.m	1	6	0.059
case24_ieee_rts.m	1	4	0.012
case2736sp.m	1	4	0.053
case2737sop.m	1	5	0.060
case2746wop.m	1	4	0.053
case2746wp.m	1	4	0.054
case2848rte.m	1	3	0.043
case2868rte.m	1	4	0.056
case2869pegase.m	1	6	0.084
case30.m	1	3	0.010
case300.m	1	5	0.019
case30Q.m	1	3	0.009
case30pwl.m	1	3	0.010
case39.m	1	1	0.008
case4_dist.m	1	3	0.010
case4gs.m	1	3	0.011
case5.m	1	3	0.011
case57.m	1	3	0.010
case60nordic.m	1	1	0.008
case6468rte.m	1	6	0.144
case6470rte.m	1	4	0.111
case6495rte.m	1	5	0.130
case6515rte.m	1	4	0.116
case6ww.m	1	3	0.010
case8387pegase.m	1	3	0.143
case89pegase.m	1	5	0.015
case9.m	1	3	0.011
case9241pegase.m	1	6	0.243
case9Q.m	1	3	0.011
case9target.m	1	4	0.010
case_ACTIVSg10k.m	1	4	0.157
case_ACTIVSg200.m	1	2	0.010
case_ACTIVSg2000.m	1	3	0.042
case_ACTIVSg25k.m	1	7	0.549
case_ACTIVSg500.m	1	3	0.015
case_ACTIVSg70k.m	1	5	1.398
case_RTS_GMLC.m	1	3	0.013
case_SyntheticUSA.m	1	5	1.727
case_ieee30.m	1	2	0.008

PSS/E Dyr Parser

ANDES supporting parsing PSS/E dynamic files in the format of .dyr. Support new dynamic models can be added by editing the input and output conversion definition file in andes/io/psse-dyr.yaml, which is in the standard YAML format. To add support for a new dynamic model, it is recommended to start with an existing model of similar functionality.

Consider a GENCLS entry in a dyr file. The entry looks like

1 'GENCLS' 1    13.0000  0.000000  /

where the fields are in the order of bus index, model name, generator index on the bus, inertia (H) and damping coefficient (D).

The input-output conversion definition for GENCLS is as follows

GENCLS:
    destination: GENCLS
    inputs:
        - BUS
        - ID
        - H
        - D
    find:
        gen:
            StaticGen:
                bus: BUS
                subidx: ID
    get:
        u:
            StaticGen:
                src: u
                idx: gen
        Sn:
            StaticGen:
                src: Sn
                idx: gen
        Vn:
            Bus:
                src: Vn
                idx: BUS
        ra:
            StaticGen:
                src: ra
                idx: gen
        xs:
            StaticGen:
                src: xs
                idx: gen
    outputs:
        u: u
        bus: BUS
        gen: gen
        Sn: Sn
        Vn: Vn
        D: D
        M: "GENCLS.H; lambda x: 2 * x"
        ra: ra
        xd1: xs

It begins with a base-level definition of the model name to be parsed from the dyr file, namely, GENCLS. Five directives can be defined for each model: destination, inputs, outputs, find and get. Note that find and get are optional, but the other three are mandatory.

• destination is ANDES model to which the original PSS/E model will be converted. In this case, the ANDES model have the same name GENCLS.
• inputs is a list of the parameter names for the PSS/E data. Arbitrary names can be used, but it is recommended to use the same notation following the PSS/E manual.
• outputs is a dictionary where the keys are the ANDES model parameter and the values are the input parameter or lambda functions that processes the inputs (see notes below).
• find is a dictionary with the keys being the temporary parameter name to store the idx of external devices and the values being the criteria to locate the devices. In the example above, GENCLS will try to find the idx of StaticGen with bus == BUS and the subidx == ID, where BUS and ID are from the dyr file.
• get is a dictionary with each key being a temporary parameter name for storing an external parameter and each value being the criteria to find the external parameter. In the example above, a temporary parameter u is the u parameter of StaticGen whose idx == gen. Note that gen is the idx of StaticGen retrieved in the above find section.

For the inputs section, one will need to skip the model name because for any model, the second field is always the model name. That is why for GENCLS below, we only list four input parameters.

1 'GENCLS' 1    13.0000  0.000000  /

For the outputs section, the order can be arbitrary, but it is recommended to follow the input order as much as possible for maintainability. In particular, the right-hand-side of the outputs can be either an input parameter name or an anonymous expression that processes the input parameters. For the example of GENCLS, since ANDES internally uses the parameter of M = 2H, the input H needs to be multiplied by 2. It is done by the following

M: "GENCLS.H; lambda x: 2 * x"

where the left-hand-side is the output parameter name (destination ANDES model parameter name), and the right-hand-side is arguments and the lambda function separated by semi-colon, all in a pair of double quotation marks. Multiple arguments are accepted and should be separated by comma. Arguments can come from the same model or another model. In the case of the same model, the model name can be neglected, namely, by writing M: "H; lambda x: 2 * x".