citylearn.dynamics module

class citylearn.dynamics.Dynamics

Bases: object

Base building dynamics model.

reset()

class citylearn.dynamics.LSTMDynamics(filepath: Path | str, input_observation_names: List[str], input_normalization_minimum: List[float], input_normalization_maximum: List[float], hidden_size: int, num_layers: int, lookback: int, input_size: int = None, dropout: float = None)

Bases: Dynamics, Module

LSTM building dynamics model that predicts indoor temperature based on partial cooling/heating load and other weather variables.

Parameters:

• filepath (Union[Path, str]) – Path to model state dictionary.

• input_observation_names (List[str]) – List of maximum values used for input observation min-max normalization.

• input_normalization_minimum (List[float]) – List of minumum values used for input observation min-max normalization.

• input_normalization_maximum (List[float]) – List of maximum values used for input observation min-max normalization.

• hidden_size (int) – The number of neurons in hidden layer.

• num_layers (int) – Number of hidden layers.

• lookback (int) – Number of samples used for prediction.

• input_size (int, optional) – Number of variables used for prediction. This may not equal input_observation_names e.g. cooling and heating demand may be included in input_observation_names but only one of two may be used for the actual prediction depending on building needs. The default is to set set input_size to the length of input_observation_names.

• dropout (float, default: 0.0) – Probability of excluding input and recurrent connections to LSTM units from activation and weight updates while training a network. This has the effect of reducing overfitting and improving model performance.

forward(x, h)

Predict indoor dry bulb temperature.

init_hidden(batch_size: int)

Initialize hidden states.

property input_size: int

reset()

Loads dynamic model state dict, and initializes hidden states and model input.

set_linear() → Linear

Initialize linear transformer.

set_lstm() → LSTM

Initialize LSTM model.

terminate()