SmartBuildSim

SmartBuildSim is a deterministic smart-building simulation toolkit. It provides utilities for loading BIM-style building schemas, generating synthetic sensor data, engineering features, training forecasting/anomaly/clustering/RL models, and producing Matplotlib visualisations. A Typer-powered CLI orchestrates the full workflow using YAML configuration files.

Features

• Deterministic synthetic data – centrally managed via smartbuildsim.config with repeatable random seeds.
• Rich modelling suite – forecasting (linear regression with lagged features), anomaly detection (IsolationForest), clustering (k-means) and reinforcement learning (tabular Q-learning).
• Scenario presets – ready-to-run building layouts with tuned defaults.
• Matplotlib visualisations – overlay anomalies and cluster assignments on sensor time series plots.
• Typer CLI – declarative YAML configs with dotted overrides and sensible output management.

Quickstart

1. Install dependencies (Python 3.10+):

   pip install -e .
   

2. Generate a default BIM schema:

   smartbuildsim bim init examples/outputs/schema.yaml --scenario office-small
   

3. Create synthetic data, train models, detect anomalies, cluster zones, train RL policy, and plot results using the bundled configuration:

   smartbuildsim data generate examples/configs/default.yaml
   smartbuildsim model forecast examples/configs/default.yaml
   smartbuildsim model anomalies examples/configs/default.yaml
   smartbuildsim cluster run examples/configs/default.yaml
   smartbuildsim rl train examples/configs/default.yaml
   smartbuildsim viz plot examples/configs/default.yaml \
     --anomalies-path outputs/anomalies.csv --clusters-path outputs/clusters.csv
   

4. Run the end-to-end Python example:

   python examples/scripts/run_example.py
   

5. Explore the interactive workflow notebook located at examples/notebooks/smartbuildsim_workflow.ipynb for a cell-by-cell walkthrough of the same pipeline. Launch it with your preferred Jupyter interface (e.g. jupyter lab examples/notebooks).

Configuration Overview

Configuration is supplied via YAML documents. The examples/configs/default.yaml file demonstrates the expected structure:

scenario: office-small
determinism:
  seed: 123
paths:
  output_dir: ./outputs
  dataset: ./outputs/dataset.csv
data:
  days: 10
  seed: 123
models:
  forecasting:
    horizon: 2
  anomaly:
    contamination: 0.07
cluster:
  sensors:
    - cluster_energy
    - cluster_co2
viz:
  sensor: office_energy

Override any configuration entry directly from the CLI using dotted keys:

smartbuildsim data generate examples/configs/default.yaml \
  --override data.seed=999 --override data.days=5

See Deterministic execution for component specific seed overrides that integrate with the CLI.

Reproducibility

Run ./reproduce.sh to execute the default CLI pipeline twice and assert that the generated datasets are identical. This lightweight check is useful both for local development and as a CI smoke test.

Development

Install the project in editable mode and run the quality gates:

pip install -e .[dev]
ruff check .
mypy .
pytest

Post-generation Checklist

• Deterministic data generation with scenario presets
• Forecasting, anomaly detection, clustering, and RL modules
• Matplotlib-based visualisations
• Typer CLI with overrides and persisted outputs
• Comprehensive tests and CI configuration

License

This project is licensed under the MIT License.

With Passion: dr Tymoteusz Miller