Command line#


ANDES is invoked from the command line using the command andes. Running andes without any input is equal to andes -h or andes --help. It prints out a preamble with version and environment information, followed by and help commands

    _           _         | Version 1.6.0
   /_\  _ _  __| |___ ___ | Python 3.9.10 on Linux, 03/12/2022 10:30:44 AM
  / _ \| ' \/ _` / -_|_-< |
 /_/ \_\_||_\__,_\___/__/ | This program comes with ABSOLUTELY NO WARRANTY.

usage: andes [-h] [-v {1,10,20,30,40}]
            {run,plot,doc,misc,prepare,prep,selftest,st,demo} ...

positional arguments:
                        [run] run simulation routine; [plot] plot
                        results; [doc] quick documentation; [misc] misc.
                        functions; [prepare] prepare the numerical code;
                        [selftest] run self test;

optional arguments:
-h, --help            show this help message and exit
-v {1,10,20,30,40}, --verbose {1,10,20,30,40}
                        Verbosity level in 10-DEBUG, 20-INFO, 30-WARNING,
                        or 40-ERROR.


If the andes command is not found, it could be due to

  1. missed steps in your installation process

  2. errors during installation

  3. forgot to activated the environment with ANDES

andes accepts an optional arugment to control verbosity level. It is done through -v LEVEL or --verbose LEVEL, where level is a number. Logging level by default is 20 (INFO) and can be chosen from:

  • 1 (DEBUG with code location info)

  • 10 (DEBUG)

  • 20 (INFO)

  • 30 (WARNING)

  • 40 (ERROR)

  • 50 (CRITICAL)

To show debugging outputs, use andes -v 10, followed by top-level commands. To only show warnings and errors, use andes -v 30.

The top-level commands are {run,plot,doc,misc,prepare,selftest}. Each command contains a group of subcommands, which can be looked up with -h. For example, use andes run -h to look up the subcommands for andes run. Frequently used commands are explained below.


Some subcommands have shorthand names:

  • andes st is equivalent to andes selftest

  • andes prep is equivalent to andes prepare

andes selftest#

After the installation, please run andes selftest from the command line to test ANDES functionality. It might take a minute to run the full self-test suite. An example output looks like

test_docs (test_1st_system.TestCodegen) ... ok
... (outputs are truncated)
test_pflow_mpc (test_pflow_matpower.TestMATPOWER) ... ok
Ran 60 tests in 10.109s


There may be more test than what is shown above. Make sure that all tests have passed.

ANDES receives frequent updates. After each update, please run andes st to confirm the functionality. The command also makes sure the generated code is up to date. See andes prepare for more details on automatic code generation.


There is a quick mode to test ANDES by skipping code generation. This should only be used when you are certain that there is no modification to models between the last code generation and now.

The quick mode is invoked by andes st -q.

andes prepare#

The symbolically defined models in ANDES need to be generated into numerical code for simulation. The code generation process is automatic the first time you use ANDES to run any case study. It takes 10 seconds to one minute to generate the code depending on your platform. When done, no code generation is needed in your future use untill you modify the models.

It is also possible to generate the code manually with andes prepare or andes prep. In addition, andes selftest automatically calls the code generation.


Generated code files are stored in Python code in $HOME/.andes/pycode. While being human-readable, they are not human-friendly and should only be consulted during low-level debugging.

The default code generation mode is known as the "quick mode". It skips the generation of \(\LaTeX\)-formatted equations, which are only useful in documentation and the interactive mode.

Option -i or --incremental can be used to speed up code generation during model development. andes prepare -i only generates code for models that are detected with changes since the last code generation.


To developers:

andes prepare -i needs to be called following model changes, such as equation modification and adding variables. Otherwise, due to mismatches in model and code, simulation results will not reflect the new changes, at best, or even lead to unexpected errors

ANDES supports precompiling the generated Python code using Numba. See Numba compilation.

andes run#

andes run is the entry point for power system analysis routines. The full list of options can be printed with andes run -h. andes run takes one positional argument, filename, along with other optional keyword arguments. filename is the path to cases, either relative or absolute.

  • Relative path: andes run kundur_full.xlsx, e.g., uses a relative path. It works only if kundur_full.xlsx exists in the working directory of the command line.

  • Absolute path: andes run /Users/hcui7/kundur_full.xlsx (on macOS) or andes run C:/Users/hcui7/kundur_full.xlsx (on Windows) use absolute paths to the case files. They do not depend on the command-line current directory.


When working with the command line, use cd to change directory to the folder containing your test case. Spaces in folder and file names need to be escaped properly, so it's generally advised to avoid spaces in file and folder names.

To find out your current working directory, look for the line below the ANDES preamble that reads like

Working directory: "/home/hacui/repos/andes/andes/cases/kundur"

Input path#

ANDES allows one to specify the path to look for the case file instead of the working directory. This is done by using the -p or --input-path option. For example, if kundur_full.xlsx is in folder /home/hacui/cases, one can do

andes run kundur_full.xlsx -p /home/hacui/cases

The argument passed to -p or --input-path can also be a relative path. If you need further help understanding paths, please consult other online articles.


ANDES takes multiple files inputs or wildcard. Multiprocessing will be triggered if more than one valid input files are passed to filename.

  • Multiple files: to run the power flow for kundur_full.xlsx and kundur_motor.xlsx simultaneously, one can do

andes run kundur_full.xlsx kundur_motor.xlsx

The output will look like

Working directory: "/home/hacui/repos/andes/andes/cases/kundur"
-> Processing 2 jobs on 12 CPUs.
Process 0 for "kundur_full.xlsx" started.
Process 1 for "kundur_motor.xlsx" started.
Log saved to "/tmp/andes/andes-uopdutii/andes.log".
-> Multiprocessing finished in 2.4680 seconds.
  • Wildcard: to run power flow for files with a prefix of kundur_ and a suffix (file extension) of .xlsx, run

andes run kundur_*.xlsx

Case files with such name pattern, including kundur_full.xlsx and kundur_motor.xlsx, among others, will be processed.

Option --ncpu NCPU can be used to specify the maximum number of parallel processes. By default, all cores will be used. A small number can be specified to increase operating system responsiveness.


Option -r or -routine is used for specifying the analysis routine, followed by the routine name. Available routine names include

  • pflow for power flow calculation

  • tds for time domain simulation

  • eig for eigenvalue analysis

If -r is not given, the power flow calculation routine will be called. There are routine specific options that can be passed to andes run and are discussed next.

Each routine has a list of configuration options (called "config") to control their behaviors. Config needs to be distinguished from command-line options as not all config options are available in the command-line. Refer to Config for details.

Power flow#


Examples in the following will utilize the kundur_full.xlsx test case. If you have cloned the ANDES repository, it can be found in andes/cases/kundur in the source code folder. You can also download it from here.

To run power flow, change to the directory containing kundur_full.xlsx, and execute the following in the command line:

andes run kundur_full.xlsx

Alternatively, the full path to the case file is also recognizable, such as

andes run /home/user/andes/cases/kundur/kundur_full.xlsx

The power flow report will be saved to the current directory where ANDES is invoked. The report contains four sections:

  1. system statistics,

  2. ac bus and dc node data

  3. ac line data, and

  4. results of other algebraic variables and state variables.

By default, the power flow routine is configured to use full Newton Raphson method, and reactive power limits are not checked. To change these config, edit the config file by referring to andes doc PFlow and andes doc PV.

Following power flow, ANDES does not initialize dynamic models to save time. When developing dynamic models, one can enable the initialization by setting in the config file

init_tds = 1

Time-domain simulation#

To run the time domain simulation (TDS) for kundur_full.xlsx, run

andes run kundur_full.xlsx -r tds

The output looks like:

Parsing input file "kundur_full.xlsx"...
Input file parsed in 0.1533 seconds.
System internal structure set up in 0.0174 seconds.
-> System connectivity check results:
No islanded bus detected.
System is interconnected.
Each island has a slack bus correctly defined and enabled.

-> Power flow calculation
Sparse solver: KLU
Solution method: NR method
Numba compilation initiated with caching.
Power flow initialized in 0.1428 seconds.
0: |F(x)| = 14.9282832
1: |F(x)| = 3.608627841
2: |F(x)| = 0.1701107882
3: |F(x)| = 0.002038626956
4: |F(x)| = 3.745103977e-07
Converged in 5 iterations in 0.0014 seconds.
Report saved to "kundur_full_out.txt" in 0.0004 seconds.

-> Time Domain Simulation Summary:
Sparse Solver: KLU
Simulation time: 0.0-20.0 s.
Fixed step size: h=33.33 ms. Shrink if not converged.
Numba compilation initiated with caching.
Initialization for dynamics completed in 0.0626 seconds.
Initialization was successful.
<Toggler 1>: Line.Line_8 status changed to 0 at t=2.0 sec.
100%|########################################| 100/100 [00:00<00:00, 241.53%/s]
Simulation completed in 0.4141 seconds.
Outputs to "kundur_full_out.lst" and "kundur_full_out.npz".
Outputs written in 0.0171 seconds.
-> Single process finished in 0.8890 seconds.

The output contains the key information for the simulation, such as solver name and step size. It prints out the disturbance information, the trip of Line Line_8 at time t=2.0 sec.

There are a few places needing to be noted:

  1. Make sure the power flow calculation is successful. Otherwise, there is no good starting point for dynamic simulation.

  2. Make sure no suspect initialization error is found. Otherwise, the system will not be at steady state even before disturbances.

TDS writes two output files: a variable list file kundur_full_out.lst, and a compressed NumPy data file kundur_full_out.npz:

  • List file: it is a plain-text file with three columns: variable indices, variable name in plain text, and variable name in the \(\LaTeX\) format. The variable indices are needed to plot the needed variable.

  • Data file: it is a zipped NumPy binary file. Although not directly editable, it can be used for plotting or can be converted to a CSV file. See the subsection on andes plot.

There are TDS-specific options that can be passed to andes run:

  • --tf TF: the final time of the simulation. TF should be a number in seconds. By default, it is set to 20.0.

  • --addfile ADDFILE: specify an additional data file. This is currently used to supply PSS/E dyr file in addition to a raw file.

  • --flat: turn on "flat run" mode to ignore all disturbances. The simulation will be performed up to the end time.

  • --no-pbar: turn off progress bar.

  • --from-csv FROM_CSV: use data from a CSV file to perform mock simulation. The CSV file should be in the format of andes plot --to-csv.

Disable output#

Output to files can be disabled with --no-output or -n. It is useful when computation is needed but results can be discarded. It is also useful when results are processed in memory, combined with the --shell option discussed next.

IPython shell#

The ANDES CLI will exit to the system shell when finished running. It is sometimes useful to script in Python to quickly process the simulation results in memory, such as plotting. ANDES can exit to the IPython shell with --shell or -s. For example:

andes run kundur_full.xlsx -r tds -s -n

Note the -n is optional to disable file output. The terminal output will look like

<Toggler 1>: Line.Line_8 status changed to 0 at t=2.0 sec.
100%|#########################################| 100/100 [00:00<00:00, 246.07%/s]
Simulation completed in 0.4064 seconds.
Outputs to "kundur_full_out.lst" and "kundur_full_out.npz".
Outputs written in 0.0171 seconds.
-> Single process finished in 0.8796 seconds.
IPython: Access System object in variable `system`.
Python 3.9.10 | packaged by conda-forge | (main, Feb  1 2022, 21:24:11)
Type 'copyright', 'credits' or 'license' for more information
IPython 8.1.1 -- An enhanced Interactive Python. Type '?' for help.

In [1]:

A prompt will appear following In [1]: to indicate an IPython shell. If the test case file is parsed without error, the system object will be stored in variable system, i.e.

In [1]: system
Out[1]: <andes.system.System at 0x7fc1cd992790>

Python commands can be executed thereafter. To exit, type exit and press enter.

Format converter#

ANDES uses the Excel format to store power system data in the ANDES semantics. In addition, multiple input formats are recognized and can be converted to the ANDES xlsx format. Converting data into the ANDES has pros and cons:

  • Pros: - The data can be readily edited with an Excel-like tool - Data for models unique to ANDES can be readily added to the xlsx file

  • Cons: - Conversion from ANDES xlsx back to the original format is not supported


It is recommended to stay with the original data format to maximize compatibility when no ANDES-specific models are used.

Format conversion is done through --convert FORMAT or -c FORMAT, where FORMAT is the output format. For now, the following formats are supported:

  • xlsx: an Excel spread sheet format with ANDES-specific data. It is not compatible with xlsx with datafrom other tools such as Pandapower.

  • json: a JSON plain-text file with ANDES-specific data. Likewise, it is unlikely to be compatible with JSON from other power system tools. JSON is much faster to parse than xlsx but not as friendly to edit.

To convert kundur_full.xlsx, for example, to the json format, run

andes run kundur_full.xlsx --convert json

The output messages will look like

Parsing input file "kundur_full.xlsx"...
Input file parsed in 0.1576 seconds.
System internal structure set up in 0.0175 seconds.
JSON file written to "kundur_full.json"
Format conversion completed in 0.0053 seconds.
-> Single process finished in 0.2582 seconds.

Note that --convert will only create sheets for existing models.

The converter works with any input formats that are currently supported. These include:

  • .m: MATPOWER case file

  • .raw and .dyr: PSS/E raw and dyr files

  • .xlsx: Excel spreadsheet file with ANDES data

  • .json: JSON plain-text file with ANDES data

PSS/E inputs#

To work with PSS/E input files (.raw and .dyr), one need to provide the raw file through casefile and pass the dyr file through --addfile. For example, in andes/cases/kundur, one can run the power flow using

andes run kundur.raw

and run a no-disturbance time-domain simulation using

andes run kundur.raw --addfile kundur_full.dyr -r tds


If one wants to modify the parameters of models that are supported by both PSS/E and ANDES, one can directly edit those dynamic parameters in the .raw and .dyr files to maintain interoperability with other tools.

To create add a disturbance, there are two options. The recommended option is to convert the PSS/E data into an ANDES xlsx file, edit it and run (see the previous subsection). The alternative approach is documented in Creating disturbances.


Profiling is useful for analyzing the computation time and code efficiency. Option --profile enables the profiling of ANDES execution. The profiling output will be written in two files in the current folder, one ending with _prof.txt and the other one with

The text file can be opened with a text editor, and the .prof file can be visualized with snakeviz, which can be installed with pip install snakeviz.

If the output is disabled, profiling results will be printed to stdio.

andes plot#

andes plot is the command-line tool for plotting. It currently supports time-domain simulation data. Three positional arguments are required, and a dozen of optional arguments are supported.

positional arguments:




simulation output file name, which should end with out. File extension can be omitted.


the X-axis variable index, typically 0 for Time


Y-axis variable indices. Space-separated indices or a colon-separated range is accepted

For the list of optional arguments, see the output of andes plot -h.

To plot the generator speed variable omega of GENROU_1 omega GENROU 1 versus time, one way is to supply the variable indices found in the .lst output file. The index of the variable omega GENROU 1 is found to be 5, and Time is found to be 0, so the plot command should be

andes plot kundur_full_out.lst 0 5

where kundur_full_out.lst is list file name, 0 is the index of Time for the x-axis, and 5 is the index of omega GENROU 1. Note that for the the file name, either kundur_full_out.lst or kundur_full_out.npz works as the program will automatically extract the file name.

The y-axis variable indices can also be specified as a Python range. For example, andes plot kundur_full_out.npz 0 2:21:6 will plot the variables with indices 2, 8, 14 and 20.

It can become tedious to look up the indices of variables in the .lst file. andes plot supports --xargs or -a for searching for variable indices and passing them as arguments to andes plot. See Examples - "Using CLI from Notebook".

LaTeX rendering#

andes plot will attempt to render with \(\LaTeX\) if dvipng program is in the search path. Figures rendered by \(\LaTeX\) has publication-quality aesthetics for symbols but takes considerably longer time. In case \(\LaTeX\) is available but fails (frequently happens on Windows), the option -d can be used to disable \(\LaTeX\) rendering.

andes doc#

andes doc is a handy tool to look up model, routine and config documentation. Model documentation include the descriptions of parameters, variables, and configs. A pretty-print version is available online in Model reference.

The basic usage of andes doc is to provide a model name or a routine name as the positional argument. For a model, it will print out model parameters, variables, and equations to the stdio. For a routine, it will print out fields in the Config file.


For full model documentation, visit Model reference.

For example, to check the parameters for model Toggler, run

$ andes doc Toggler
Model <Toggler> in Group <TimedEvent>

    Time-based connectivity status toggler.


 Name  |         Description          | Default | Unit |    Type    | Properties
 u     | connection status            | 1       | bool | NumParam   |
 name  | device name                  |         |      | DataParam  |
 model | Model or Group of the device |         |      | DataParam  | mandatory
       | to control                   |         |      |            |
 dev   | idx of the device to control |         |      | IdxParam   | mandatory
 t     | switch time for connection   | -1      |      | TimerParam | mandatory
       | status                       |         |      |            |

To list all supported models, run

$ andes doc -l
Supported Groups and Models

     Group       |                   Models
 ACLine          | Line
 ACTopology      | Bus
 Collection      | Area
 DCLink          | Ground, R, L, C, RCp, RCs, RLs, RLCs, RLCp
 DCTopology      | Node
 Exciter         | EXDC2
 Experimental    | PI2
 FreqMeasurement | BusFreq, BusROCOF
 StaticACDC      | VSCShunt
 StaticGen       | PV, Slack
 StaticLoad      | PQ
 StaticShunt     | Shunt
 SynGen          | GENCLS, GENROU
 TimedEvent      | Toggler, Fault
 TurbineGov      | TG2, TGOV1

To view the Config fields for a routine, run

$ andes doc TDS
Config Fields in [TDS]

  Option   | Value |                  Info                  | Acceptable values
 sparselib | klu   | linear sparse solver name              | ('klu', 'umfpack')
 tol       | 0.000 | convergence tolerance                  | float
 t0        | 0     | simulation starting time               | >=0
 tf        | 20    | simulation ending time                 | >t0
 fixt      | 0     | use fixed step size (1) or variable    | (0, 1)
           |       | (0)                                    |
 shrinkt   | 1     | shrink step size for fixed method if   | (0, 1)
           |       | not converged                          |
 tstep     | 0.010 | the initial step step size             | float
 max_iter  | 15    | maximum number of iterations           | >=10

andes misc#

andes misc contains miscellaneous functions, such as configuration and output cleaning.


ANDES uses a configuration file to set runtime configs for the system routines, and models. andes misc --save-config saves all configs to a file. By default, it saves to $HOME/.andes/andes.conf file, where $HOME is the path to your home directory.

With andes misc --edit-config, you can edit ANDES configuration handy. The command will automatically save the configuration to the default location if not exist. The shorter version --edit can be used instead as Python matches it with --edit-config.

You can pass an editor name to --edit, such as --edit vim. If the editor name is not provided, it will use the following defaults: - Microsoft Windows: notepad. - GNU/Linux: the $EDITOR environment variable, or vim if not exist.

For macOS users, the default is vim. If not familiar with vim, you can use nano with --edit nano or TextEdit with --edit "open -a TextEdit".


andes misc -C, --clean

Option to remove any generated files. Removes files with any of the following suffix: _out.txt (power flow report), _out.npy (time domain data), _out.lst (time domain variable list), and _eig.txt (eigenvalue report).


Check the version of ANDES and the core packages it uses, use

andes misc --version

Please include the output in your bug report.