Skip to main content

Learn Thermal Model

This module defines the LearnThermalDynamics class, which manages the learning, validation, and persistence of thermal models for space heating.

This class is responsible for:

  • Retrieving historical data (indoor temperature, heater consumption, weather) from the Core API.
  • Preprocessing this data for use in a thermal model learning algorithm.
  • Executing the learning process to derive a state-space thermal model.
  • Validating if an existing thermal model is up-to-date or if re-learning is required.
  • Saving and loading learned thermal models to/from a JSON file for persistence.
  • Providing a default thermal model if learning fails or insufficient data is available.

Classes​

LearnThermalDynamics​

Manages the learning, validation, and persistence of thermal models for space heating.

Methods​

__init__()​

Initializes the LearnThermalDynamics by retrieving information about all devices.

validate_or_learn_model(start: datetime, stop: datetime) -> Dict[str, Any]​

Validates if a thermal model exists and is recent; otherwise, it learns a new one.

This method checks for a saved thermal model. If found, it verifies if the model was learned within a learning_threshold (e.g., 1 day). If the model is missing, outdated, or cannot be loaded, it triggers the execute_learning process.

Args:

  • start: The start datetime for retrieving historical data if learning is needed.
  • stop: The stop datetime for retrieving historical data if learning is needed.

Returns:

  • A dictionary containing the validated, newly learned, or default thermal model.
execute_learning(start: datetime, stop: datetime, old_model_exists: bool = True) -> Dict[str, Any]​

Executes the learning process for the thermal models.

This method retrieves historical data, processes it, and then attempts to learn a black-box thermal model using the ThermalModels class. If learning fails (e.g., due to insufficient data or optimization issues), it provides either an existing old model or a default model.

Args:

  • start: The start datetime for retrieving historical data.
  • stop: The stop datetime for retrieving historical data.
  • old_model_exists: A boolean indicating if an old model was found, used for logging purposes.

Returns:

  • A dictionary containing the learned thermal model parameters (Ax, Au, Aw) and metadata, or a default model if learning is unsuccessful.