Abstract

Distributed energy resources (DERs) are expected to provide increasingly large amounts of energy and ancillary services to the grid. However, modelling distribution-connected assets in transmission studies is a challenge due to their sheer number, and due to the low visibility in distribution grids. To tackle this challenge, one of the main approaches consists in reducing detailed models of distribution grids into equivalent models, but few researchers considered the fact that detailed models are subject to many uncertainties. In this work, we proposed to derive equivalents based on quantiles of the behaviour of detailed models. Such equivalents can then be used to obtain statistics-informed bounds on the results of any transmission studies. We demonstrate the accuracy of this approach by comparing it to probabilistic transmission and distribution (\TD{}) simulations in both critical clearing time computations and simulations of cascading outages.

Download full text here

Citation

@INPROCEEDINGS{ISGT2023_ADN,
    author={Sabot, Frédéric and Henneaux, Pierre and Lamprianidou, Ifigeneia S. and Papadopoulos, Panagiotis N.},
    booktitle={2023 IEEE PES Innovative Smart Grid Technologies Europe (ISGT EUROPE)},
    title={Statistics-informed bounds for active distribution network equivalents subject to large disturbances},
    year={2023},
    month = 10,
    doi={10.1109/ISGTEUROPE56780.2023.10408629}
}

Supplementary data

Note: The links below provide with the data that were used to write the above publication. For more up-to-date versions, please refer to the master branch of the linked repositories.

Test systems:

  • CIGRE_MV_Wind_voltage: Files to derive dynamic equivalents with only short-circuits in the training set
  • CIGRE_MV_Wind_frequency: Also considering frequency ramps in the training set
  • CIGRE_MV_Wind_infbus: Training using an infinite bus with variations (leads to an equivalent with poor performance)
  • CIGRE_MV_Wind_reduced: Leads to a simple equivalent made of a simple restorative load model
  • IEEE39_CCT: Files for critical clearing time computations on the IEEE 39-bus system (has first to be merged with a load model using the MergeTD.py script)
  • IEEE39_protected_N_2: Files to cascading outage simulations of the IEEE 39-bus system (has first to be merged with a load model using the MergeTD.py script)

Note that the above networks include dynamic models that are available here

Scripts. Python scripts used to derive the dynamic equivalents and to build the T&D networks. For more details, please refer to the documentation of the .py files or contact me.

As briefly discussed in the paper, we also tried to complement the differential evolution (DE) algorithm using a least absolute shrinkage and selection operator (LASSO) and by iteratively adding disturbances to the training set as proposed in the thesis of Gilles Chaspierre, but this led to slower convergence in our case. The scripts used for this are available in a separate branch.