Overview

Power flow linearization approaches should be readily accessible as fundamental tools. However, most existing methods, particularly the promising data-driven techniques, are unfortunately not open-sourced, nor available in existing toolkits or software. We thus developed Daline, a data-driven power flow linearization toolbox designed for power systems researchers, educators, and engineers. This free, open-source package consists of MATLAB-language M-files, featuring 57 linearization methods, including 53 data-driven methods and 4 commonly used physics-driven methods.

Functionality

(1) Data Generation. (2) Data Pollution. (3) Data Cleaning. (4) Data Normalization. (5) Method Selection. (6) Method Customization. (7) Model Linearization. (8) Model Evaluation. (9) Result Visualization.

Step 1: Download Daline

Click daline116.zip to download the latest release version, and extract the downloaded ZIP file anywhere you like.

Step 2: Install Daline

Start MATLAB and change your working directory to the Daline directory you just extracted (the one containing daline_setup.m). Daline will be installed by typing the following command in the MATLAB command window:

 daline_setup

The above command will install Daline with a quick test of its major functionality. If you want to install Daline while testing all its functionality, type the following command:

 daline_setup('full')

Step 3: Enjoy

There is no step 3, but you can now choose to verify all the functionality of Daline by typing:

 daline_test

Or to verify any specific built-in linearization methods, e.g., 'PLS_REC' and 'RR_KPC', in Daline by typing:

 daline_test({'PLS_REC'; 'RR_KPC'})

And, if you find Daline helpful for your work, please remember to cite the following reference!

@article{Daline,
  author = {Jia, Mengshuo and Chan, Wen Yi and Hug, Gabriela},
  title = {Daline: A Data-driven Power Flow Linearization Toolbox for Power Systems Research and Education}, 
  journal = {ETH Research Collection},
  year = {2024}, 
  url = {https://doi.org/10.3929/ethz-b-000681867},
  doi = {10.3929/ethz-b-000681867}
}

Quickly get a linear model

To train a linear power flow model for the standard IEEE-118 case using the default linearization method and options, type:

model = daline.all('case118')

There are over 300 options/parameters to customize the behavior of the wrapper daline.all, e.g.,

model = daline.all('case118', 'data.baseType', 'Random', 'method.name', 'LS_COD');

Or

opt = daline.setopt('data.baseType', 'Random', 'method.name', 'LS_COD');
model = daline.all('case118', opt);

The wrappers in the following examples also offer numerous options and parameters that can be adjusted using name-value pairs. These examples demonstrate only a small selection of these options to provide an easy starting point.

Generate data

To generate artificial power flow data for a user-owned MPC-structured power system using a time series loading curve, type:

data = daline.data('case.mpc', mpc_user, 'data.baseType', 'TimeSeriesRand')

There are more than 45 options available for customizing data generation.

Train & test linear model

To train a linear power flow model using an existing dataset with ridge regression and k-plane clustering:

model = daline.fit(data, 'method.name', 'RR_KPC')

Customize linearization methods

To customize ridge regression with k-plane clustering, for example by enabling the auto-tuning of the regularization factor in this method:

model = daline.fit(data, 'method.name', 'RR_KPC', 'RR.lambdaInterval', [0:1e-3:0.02])

Ridge regression with k-plane clustering has 9 hyperparameters/options for customization in Daline.

Accuracy Comparison & Visualization

To compare and visualize the linearization error distributions of various data-driven and physics-driven approaches:

models = daline.rank(data, {'PLS_SIM', 'RR', 'LS_COD', 'DC', 'TAY'});

The comparison settings and the parameters for the methods can all be adjusted using daline.rank.

Efficiency Comparison & Visualization

To compare and visualize the computational efficiency of various data-driven and physics-driven approaches:

time = daline.time(data, {'PLS_SIM', 'RR', 'LS_COD', 'DC', 'TAY'} , 'PLOT.repeat', 5);

The comparison settings and the parameters for the methods can all be adjusted using daline.time.