powergrid_synth.distribution.distribution_input_model

Input model for the Schweitzer feeder generator.

Uses a 3-D kernel density estimate (KDE) over (N_nodes, P_load, P_gen) to sample realistic combinations of feeder size and loading, as described in Section III-A of Schweitzer et al. (2017).

Module Contents

class powergrid_synth.distribution.distribution_input_model.DistributionInputModel(bw_method=None, seed=None)[source]

3-D KDE input model for feeder size and loading.

The model is fitted on reference feeders and then sampled to produce realistic (n_nodes, total_load, total_gen) triples.

Parameters:
  • bw_method (str or float, optional) – Bandwidth method passed to scipy.stats.gaussian_kde.

  • seed (int or None, optional) – Random seed for reproducibility.

_fit_kde(data)[source]

Fit KDE with regularization for near-singular covariance.

Parameters:

data (numpy.ndarray)

Return type:

scipy.stats.gaussian_kde

fit(feeders)[source]

Fit the KDE from a collection of reference feeder graphs.

Each graph must have node attributes P_mw (positive for load, negative for generation).

Parameters:

feeders (sequence of nx.Graph) – Reference feeder graphs.

Return type:

self

fit_from_arrays(n_nodes, total_load, total_gen)[source]

Fit directly from arrays (no graph objects needed).

Parameters:
  • n_nodes (array-like of int)

  • total_load (array-like of float)

  • total_gen (array-like of float)

Return type:

self

pdf(n_nodes, total_load, total_gen)[source]

Evaluate the KDE density at a given point.

Parameters:
  • n_nodes (float) – Coordinates in the 3-D input space.

  • total_load (float) – Coordinates in the 3-D input space.

  • total_gen (float) – Coordinates in the 3-D input space.

Returns:

Estimated probability density.

Return type:

float

sample(n_samples=1)[source]

Draw samples from the fitted input model.

Parameters:

n_samples (int) – Number of samples to draw.

Return type:

list of FeederInputSample

class powergrid_synth.distribution.distribution_input_model.FeederInputSample[source]

A single draw from the 3-D input model.