sustaingym.envs.building.stochastic_generator#

Module Contents#

Classes#

StochasticUncontrollableGenerator

A generator class for the uncontrollable features in BuildingEnv.
class sustaingym.envs.building.stochastic_generator.StochasticUncontrollableGenerator(block_size: int)[source]#

A generator class for the uncontrollable features in BuildingEnv.

Parameters:

block_size (int) –

split_observations_into_seasons(data: numpy.ndarray) → tuple[numpy.ndarray, numpy.ndarray][source]#

Splits observation data into summer and winter seasons.

Parameters:

data (numpy.ndarray) – Observation data, shape [n, num_features]. Assumed to represent a whole year of observations.

Returns:

  • summer_observations – Summer ambient features. Shape is (len_of_season, num_features).

  • winter_observations – Winter ambient features. Shape is (len_of_season, num_features).

Raises:

ValueError if no data is given and the class instance – has no observations stored.

Return type:

tuple[numpy.ndarray, numpy.ndarray]

get_empirical_dist(season: str | None = None, this_season_observations: numpy.ndarray | None = None, block_size: int | None = None) → list[scipy.stats.rv_continuous][source]#

Fits a multivariate normal distribution to each ambient feature.

Parameters:

  • season (str | None) – The desired season. Can be None if this_season_observations is not None. Otherwise, can be summer or winter if data has been generated and split into seasons through split_observations_into_seasons.

  • this_season_observations (numpy.ndarray | None) – The observation data for this season. Can be None if user has generated and split data.

  • block_size (int | None) – Desired block size for each ambient feature vector, defaults to self.block_size

Returns:

empirical_dists – list of length num_features, empirical distributions for each of the ambient features, each distribution is a multivariate normal with shape [block_size]

Raises:

ValueError if neither season nor this_season_observations is specified OR – if the season is given but it is not “summer” or “winter”

Return type:

list[scipy.stats.rv_continuous]

draw_samples_from_dist(num_samples: int, summer_frac: float, dists: collections.abc.Iterable[scipy.stats.rv_continuous] | None = None, block_size: int | None = None) → numpy.ndarray[source]#

Draw vector samples from fitted multivariate Gaussian.

Parameters:

  • num_samples (int) – The number of desired samples.

  • summer_frac (float) – The weight of the generated observations to be given to those generated from the summer distribution.

  • dists (collections.abc.Iterable[scipy.stats.rv_continuous] | None) – Iterable of the empirical distributions. Can be None if instance has generated empirical distributions through get_empirical_dist.

  • block_size (int | None) –

Returns:

samples – The samples generated from the fitted distributions. Shape is (num_samples x block_size, num_obs_features).

Raises:

ValueError if summer_frac is not between 0 and 1 or if – either the summer or winter distributions aren’t available to draw samples from.

Return type:

numpy.ndarray