junbozhao / ac-cfm

AC cascading failure model based on MATPOWER for resilience analysis of power networks.

AC-CFM

AC cascading failure model based on MATPOWER for resilience analysis of power networks.

Licensing and Citing

We request that publications derived from the use of AC-CFM explicitly acknowledge that fact by citing the following publication:

Noebels, M., Preece, R., Panteli, M. "An AC Cascading Failure Model for Resilience Analysis in Power Networks," IEEE Systems Journal (accepted).

Getting Started

The following steps will guide you through getting AC-CFM run on your computer.

Prerequisites

• Matlab 2019a or later is recommended to run AC-CFM.

• Matpower 7.1 or later is required to run AC-CFM. Please follow the instructions on the Matpower Website for installation.

• IPOPT is recommended to install, as it tends to have higher convergence for AC optimal power flows. Instructions on how to install IPOPT can be found on the Matpower website. The OPTI Toolbox is a simple way of installing IPOPT.

• Test the installation of Matpower and IPOPT by running the following command in Matlab:

runopf(case9, mpoption('opf.ac.solver', 'IPOPT'))

Installation

1. Clone the repository

git clone https://github.com/mnoebels/AC-CFM.git

2. Add the AC-CFM folder to your Matlab path setting (menu Home -> Set Path)

Usage

Single contingency

Use the following example to model the cascade following a single contingency (here: failure of line 9 in the IEEE 39-bus test network):

% load default settings
settings = get_default_settings();
 
% enable verbose output – just for testing
settings.verbose = 1;
 
% model outage of line 9; this can also be an array of branch indices
initial_contingency = 9;
 
% apply the model
result = accfm(case39, struct('branches', initial_contingency), settings);

This should produce the following output:

  Demand increased by 0.1% (limit is 15.0%) and generation capacity is met. Distribute slack generation.
  Exceeded line ratings: 6-11

  Demand increased by 0.5% (limit is 15.0%) and generation capacity is met. Distribute slack generation.
  Q outside limits at generators at buses 34
  Exceeded line ratings: 3-4 3-18 10-13 13-14 14-15 16-17 17-18

   3 islands and 2 isolated nodes detected
   Island: [ 1 2 3 4 5 6 7 8 9 17 25 26 27 28 29 30 31 37 38 39 ]
    Demand increased by 20.9% (limit is 15.0%) or generation capacity is not met. Perform underfrequency load shedding of 12.2%.

    Q outside limits at generators at buses 31 39
    Voltage outside limits at buses 4 7 8
    Undervoltage load shedding applied at buses 4 7 8
    Exceeded line ratings: 1-2 8-9 9-39

     3 islands and 1 isolated nodes detected
     Island: [ 2 3 17 25 26 27 28 29 30 37 38 ]
      Demand decreased by 42.6% (limit is 15.0%). Tripping 2 smallest generators.

      Loads shed (22.54%) due to voltage collapse at buses 26 29

      Demand increased by 0.0% (limit is 15.0%) and generation capacity is met. Distribute slack generation.
      Exceeded line ratings: 2-25 2-30

        2 islands and 1 isolated nodes detected
        Island: [ 17 25 26 27 28 29 37 38 ]
         No generation available.

        Island: [ 2 3 ]
         No generation available.

        Island: [ 30 ]
         Demand decreased by 100.0% (limit is 15.0%). Tripping 1 smallest generators.

     Island: [ 4 5 6 7 8 31 ]
      Demand increased by 400.4% (limit is 15.0%) or generation capacity is not met. Perform underfrequency load shedding of 79.0%.


     Island: [ 1 39 ]
      Demand decreased by 4.0% (limit is 15.0%). Distribute slack generation.

     Island: [ 9 ]
      No generation available.

   Island: [ 15 16 19 20 21 22 23 24 33 34 35 36 ]
    Demand decreased by 8.4% (limit is 15.0%). Distribute slack generation.

   Island: [ 10 11 12 13 32 ]
    Demand decreased by 98.7% (limit is 15.0%). Tripping 1 smallest generators.

    No generation available.

   Island: [ 14 ]
    No generation available.

   Island: [ 18 ]
    No generation available.

Cascade halted. Elapsed time: 6.51s
Total load shedding: 45.05%
Load shedding UFLS: 20.97% 
Load shedding UVLS: 0.88% 
Load shedding VCLS: 4.61% 
Load shedding non-converging OPF: 0.00% 
Load shedding tripped: 18.59%

Batch processing

AC-CFM comes with two routines that can be used for batch processing of large numbers of contingencies, depending if the contingencies are known, or whether they should be sampled from a probability distribution. Both functions make use of the parallel processing capabilities of Matlab using parfor loops.

If the contingencies are known, the following code runs AC-CFM on the specified network for every contingency specified in scenarios. scenarios is a cell array.

result = accfm_branch_scenarios(network, scenarios, settings)

It returns a struct containing the results, tripped buses, lines, generators, etc. for each contingency.

If contingencies should be sampled from a probability distribution, use the following code. pdf is the name of a probability distribution (at the moment, only "zipf" is implemented). alpha is the exponent of the Zipf distribution. number_of_scenarios specified the number of contingencies to be sampled. If output_file is specified, the results are saved in a file with the given filename.

result = accfm_pdf_batch(network, pdf, alpha, number_of_scenarios, settings, output_file)

Settings

Behavior of AC-CFM can be adjusted via the settings struct.

Load default settings: settings = get_default_settings();

The following options are available:

• verbose (0 or 1): If set to 0, suppress model output.

• mpopt: Matpower options struct. Please refer to the Matpower docs for details.

• max_recursion_depth (integer): Maximum recursion depth. An error is thrown if this is reached and the cascade has still not come to an end.

• uvls_per_step (0 to 1): Ratio of load that is shed per UVLS step.

• uvls_max_steps (integer): Maximum number of UVLS steps before all loads at a bus are tripped.

• dP_limit (0 to 1): Maximum generation imbalance before UFLS is applied.

• P_overhead (0 to 1): Ratio of generation overhead when applying UFLS, mainly to meet transmission losses.

• Q_tolerance (0 to 1): Ratio of reactive power limits that can be exceeded before O/UXL is applied.

• grid_forming (cell array): Generator types (see Matpower function gentypes) that have grid-forming capability. Every island needs to have at least one grid-forming generator. Requires that the network struct contains a field gentypes. Set to empty to ignore requirement for grid-forming generators.

• keep_networks_after_cascade (0 or 1): In batch processing, keep final network struct for each contingency. This significantly increases required memory.

Troubleshooting

If you run into problems, for instance exceptionally large load shedding or large amount of OPF load shedding, try the following:

1. Set PMIN (column 10) of the gen matrix to zero to disable minimum power output limits of generators.
2. Increase the reactive power limits (columns 4 and 5) of the gen matrix (e.g. to +80% and -40% of PMAX).
3. Loads should be fixed (not dispatchable) before passed to AC-CFM. You can use the disp2load function provided to convert dispatchable loads to fixed loads.