DSS Python: Unofficial bindings for EPRI's OpenDSS

Python bindings and misc tools for using OpenDSS (EPRI Distribution System Simulator). Based on CFFI and DSS C-API, aiming for full COM compatibility on Windows, Linux and MacOS.

See also the other projects from DSS-Extensions.org:

• DSS C-API library: the base library that exposes a slightly modified version of EPRI's OpenDSS through a more traditional C interface, built with the open-source Free Pascal compiler instead of Delphi.
• OpenDSSDirect.py: if you don't need COM compatibility, or just would like to check its extra funcionalities. You can mix DSS Python and OpenDSSDirect.py -- for example, if you have old code using the official COM objects, you could quickly switch to DSS Python with very few code changes, and then use opendssdirect.utils to generate some DataFrames.
• OpenDSSDirect.jl: a Julia module, created by Tom Short (@tshort), recently migrated with the help of Dheepak Krishnamurthy (@kdheepak) to DSS C-API instead of the DDLL.
• DSS Sharp: available for .NET/C#, also mimics the COM classes, but Windows-only at the moment. Soon it will be possible to use it via COM too.
• DSS MATLAB: presents multi-platform integration (Windows, Linux, MacOS) with DSS C-API and is also very compatible with the COM classes.

Version 0.10.6, based on OpenDSS revision 2909 (around OpenDSS v9.0.0.3). While we plan to add a lot more funcionality into DSS Python, the main goal of creating a COM-compatible API has been reached. If you find an unexpected missing feature, please report it!

This module mimics the COM structure (as exposed via win32com or comtypes), effectively enabling multi-platform compatibility at Python level. Most of the COM documentation can be used as-is, but instead of returning tuples or lists, this modules returns/accepts NumPy arrays for numeric data exchange.

The module depends on CFFI, NumPy and, optionally, SciPy.Sparse for reading the sparse system admittance matrix.

Brief release history

• 2020-07-31 / version 0.10.6: Maintenance release to match DSS C-API 0.10.6, based on on OpenDSS revision 2909. New important settings: DSS.LegacyModels and DSS.Error.ExtendedErrors.
• 2020-03-03 / version 0.10.5: Maintenance release to match DSS C-API 0.10.5, based on on OpenDSS revision 2837. Temporarily drops the v8 parallel-machine functions, as well as conda packages on Windows.
• 2019-11-16 / version 0.10.4: Maintenance release to match DSS C-API 0.10.4.
• 2019-05-22 / version 0.10.3: Some important fixes, better general performance, new API extensions, new features ported from COM and the OpenDSS version 8 codebase.
• 2019-02-28 / version 0.10.2: Some small fixes, adds the missing CtrlQueue.Push, faster LoadShapes and new property DSS.AllowEditor to toggle editor calls.
• 2019-02-17 / version 0.10.1: Integrate DSS C-API changes/fix, some small fixes, and more error-checking.
• 2018-11-17 / version 0.10.0: Lots of changes, fixes and new features. Check the new changelog document for a list.
• 2018-08-12 / version 0.9.8: Reorganize modules (v7 and v8), adds 8 missing methods and new backend methods for OpenDSSDirect.py v0.3+. Integrates many fixes from DSS_CAPI and the upstream OpenDSS.
• 2018-04-30 / version 0.9.7: Fix some of the setters that used array data.
• 2018-04-05 / version 0.9.6: Adds missing ActiveCircuit.CktElements[index] (or ...CktElements(index)) and ActiveCircuit.Buses[index] (or ...Buses(index)).
• 2018-03-07 / version 0.9.4: Allows using len on several classes, fixes DSSProperty, and includes COM helpstrings as docstrings. Contains changes up to OpenDSS revision 2152.
• 2018-02-16 / version 0.9.3: Integrates COM interface fixes from revision 2136 (First Next iteration on some elements)
• 2018-02-12 / version 0.9.2: Experimental support for OpenDSS-PM (at the moment, a custom patch is provided for FreePascal support) and port COM interface fixes (OpenDSS revision 2134)
• 2018-02-08 / version 0.9.1: First public release (OpenDSS revision 2123)

Recent changes

Changes in 0.10.6, since 0.10.5

Check the changelog document for a detailed list for all releases.

• Updated to DSS C-API 0.10.6, which includes most changes up to OpenDSS v9.0.0.3.
• Debug builds of DSS C-API are now included. See the Debugging document.
• New DSS.LegacyModels: allow using the legacy/deprecated models for PVsystem, Storage, InvControl, and StorageController.
• New DSS.Error.ExtendedErrors: controls if the new extended error messages are used.
• Many new properties and functions in DSS.ActiveCircuit.PDElements.
• Now most of the low-level API calls are checked, mapping the errors from the DSS.Error interface to Python exceptions more frequently.

DSS C-API 0.10.6 changes:

• This version should be fully API compatible with 0.10.3+. The behavior of some functions changed with the new extensions. Especially, empty strings are explicitely return as nulls instead of "\0". This conforms to the behavior already seen in arrays of strings.
• The binary releases now use Free Pascal 3.2.0. We observed the solution process is around 6% faster, and results are even closer to the official OpenDSS.
• The releases now include both the optimized/default binary and a non-optimized/debug version. See the Debugging document for more.
• Extended API validation and Extended Errors mechanism:
  • The whole API was reviewed to add basic checks for active circuit and element access.
  • By default, invalid accesses now result in errors reported through the Error interface. This can be disabled to achieve the previous behavior, more compatible with the official COM implementation -- that is, ignore the error, just return a default/invalid value and assume the user has handled it.
  • The mechanism can be toggled by API functions DSS_Set_ExtendedErrors and DSS_Get_ExtendedErrors, or environment variable DSS_CAPI_EXTENDED_ERRORS=0 to disable (defaults to enabled state).
• New Legacy Models mechanism:
  • OpenDSS 9.0+ dropped the old PVsystem, Storage, InvControl, and StorageController models, replacing with the new versions previously known as PVsystem2, Storage2, InvControl2 and StorageController2.
  • The behavior and parameters from the new models are different -- they are better, more complete and versatile models. Check the official OpenDSS docs and examples for further information.
  • The implementation of the new models in DSS C-API was validated successfully with all test cases available. As such, we mirror the decision to make them the default models.
  • As an extension, we implemented the Legacy Models option. By toggling it, a clear command will be issued and the alternative models will be loaded. This should allow users to migrate to the new version but, if something that used to work with the old models stopped working somehow, the user can toggle the old models. The idea is to keep reproducibility of results while we keep updating the engine and the API.
  • Since EPRI dropped/deprecated the old models, we might drop them too, in a future release. Please open an issue on GitHub or send a message if those old models are important to you.
  • The mechanism can be controlled by API functions DSS_Set_LegacyModels and DSS_Get_LegacyModels, or environment variable DSS_CAPI_LEGACY_MODELS=1 to enable (defaults to disabled state).
• WireData API: expose the CapRadius property as a new pair of functions.
• PDElements API: extended with many batch functions exposing equivalents to some CSV reports: AllNames, AllMaxCurrents, AllPctNorm, AllPctEmerg, AllCurrents, AllCurrentsMagAng, AllCplxSeqCurrents, AllSeqCurrents, AllPowers, AllSeqPowers, AllNumPhases, AllNumConductors, AllNumTerminals.
• CktElement_Get_SeqPowers: fix issue for positive sequence circuits (wrong results could corrupt memory).
• Many API functions were optimized to avoid unnecessary allocations and copies.
• Some bugs found in DSS C-API and also reported upstream (already fixed in SVN):
  • CapRadius DSS property: if the radius was initialized using GMRac, CapRadius was left uninitialized, resulting in invalid/NaN values.
  • Sensors API: some functions edited capacitors instead of sensors.
• Updated to the official OpenDSS revision 2903, corresponding to versions 9.0.0+. Changes include:
  • ExportCIMXML: updated.
  • Relay: Fix in GetPropertyValue.
  • Line: In DumpProperties and MakePosSequence, the length is handled differently for lines with LineGeometry or LineSpacing.
  • Bus API: new LineList, LoadList functions.
  • Lines API: SeasonRating now returns NormAmps if there's no SeasonSignal.
  • New command DSS Zsc012: "Returns symmetrical component short circuit impedances Z0, Z1, and Z2 for the ACTIVE 3-PHASE BUS. Determined from Zsc matrix."
  • PVsystem2, Storage2, InvControl2, StorageController2 updated and renamed.

Missing features and limitations

Most limitations are inherited from dss_capi, i.e., these are not implemented:

• DSSEvents from DLL/ImplEvents.pas: seems too dependent on COM.
• DSSProgress from DLL/ImplDSSProgress.pas: would need a reimplementation depending on the target UI (GUI, text, headless, etc.).

In general, the DLL from dss_capi provides more features than both the official Direct DLL and the COM object.

Extra features

Besides most of the COM methods, some of the unique DDLL methods are also exposed in adapted forms, namely the methods from DYMatrix.pas, especially GetCompressedYMatrix (check the source files for more information).

Since no GUI components are used in the FreePascal DLL, we are experimenting with different ways of handling OpenDSS errors. Currently, the DSS.Text.Command call checks for OpenDSS errors (through the DSS.Error interface) and converts those to Python exceptions. Ideally every error should be converted to Python exceptions, but that could negatively impact performance. You can manually trigger an error check by calling the function CheckForError() from the main module.

Installing

On all major platforms, you can install directly from pip:

    pip install dss_python

Or, if you're using the Anaconda distribution, you can use:

    conda install -c pmeira dss_python

Binary wheels are provided for all major platforms (Windows, Linux and MacOS) and many combinations of Python versions (2.7, 3.5 to 3.7). If you have issues with a specific version, please open an issue about it. Conda packages support at least Python 2.7, 3.5, 3.6 and 3.7.

After a successful installation, you can then import the dss module from your Python interpreter.

Building

Get the repository:

    git clone https://github.com/dss-extensions/dss_python.git

Assuming you successfully built or downloaded the DSS C-API DLLs (check its repository for instructions), keep the folder organization as follows:

dss_capi/
dss_python/
electricdss-src/

Open a command prompt in the dss_python subfolder and run the build process:

python setup.py build
python setup.py install

If you are familiar with conda-build, there is a complete recipe to build DSS C-API, KLUSolve and DSS Python in the conda subfolder.

Example usage

If you were using win32com in code like:

import win32com.client 
dss_engine = win32com.client.gencache.EnsureDispatch("OpenDSSEngine.DSS")

or comtypes:

import comtypes.client
dss_engine = comtypes.client.CreateObject("OpenDSSEngine.DSS")

you can replace that fragment with:

import dss
dss_engine = dss.DSS

If you need the mixed-cased handling (that is, you were not using early bindings with win32com), add a call to dss.use_com_compat().

Assuming you have a DSS script named master.dss, you should be able to run it as shown below:

import dss
dss_engine = dss.DSS

dss_engine.Text.Command = "compile c:/dss_files/master.dss"
dss_engine.ActiveCircuit.Solution.Solve()
voltages = dss_engine.ActiveCircuit.AllBusVolts

for i in range(len(voltages) // 2):
    print('node %d: %f + j%f' % (i, voltages[2*i], voltages[2*i + 1]))

If you want to play with the experimental OpenDSS-PM interface (from OpenDSS v8), it is installed side-by-side and you can import it as: -- temporarily disabled in DSS Python 0.10.5. Check back in a few months.

import dss.v8
dss_engine = dss.v8.DSS

Although it is experimental, most of its funcionality is working. Depending on your use-case, the parallel interface can be an easy way of better using your machine resources. Otherwise, you can always use general distributed computing resources via Python.

Beware the v8 alternative can present issues and it should be removed as soon as all OpenDSS 8+ features are integrated into the default version.

Testing

Since the DLL is built using the Free Pascal compiler, which is not officially supported by EPRI, the results are validated running sample networks provided in the official OpenDSS distribution. The only modifications are done directly by the script, removing interactive features and some other minor issues. Most of the sample files from the official OpenDSS repository are used for validation.

The validation scripts is tests/validation.py and requires the same folder structure as the building process. You need win32com to run it on Windows.

As of version 0.11, the full validation suite can be run on the three supported platforms. This is possible by saving the official COM DLL output and loading it on macOS and Linux. We hope to automate this validation in the future.

Roadmap

Besides bug fixes, the main funcionality of this library is mostly done. Notable desirable features that may be implemented are:

• More and better documentation
• Plotting and reports integrated in Python.
• Parallel-machine properties (disabled in DSS Python 0.10.5, to be reworked).

Expect news about these items by version 0.11.

Questions?

If you have any question, feel free to open a ticket on GitHub, or contact directly me through email (pmeira at ieee.org). Please allow me a few days to respond.

Credits / Acknowlegement

DSS Python is based on EPRI's OpenDSS via the dss_capi project, check its licensing information.

This project is licensed under the (new) BSD, available in the LICENSE file. It's the same license OpenDSS uses (OPENDSS_LICENSE). OpenDSS itself uses KLUSolve and SuiteSparse, licensed under the GNU LGPL 2.1.

I thank my colleagues at the University of Campinas, Brazil, for providing feedback and helping me test this module.