https://github.com/bbartling/BrickModelSummarizer
Concept Idea to run a compressed local LLM locally on CPU
https://github.com/bbartling/BrickModelSummarizer
Last synced: about 1 year ago
JSON representation
Concept Idea to run a compressed local LLM locally on CPU
- Host: GitHub
- URL: https://github.com/bbartling/BrickModelSummarizer
- Owner: bbartling
- License: mit
- Created: 2023-07-29T19:21:19.000Z (almost 3 years ago)
- Default Branch: develop
- Last Pushed: 2024-11-01T18:59:08.000Z (over 1 year ago)
- Last Synced: 2024-11-01T19:32:04.544Z (over 1 year ago)
- Language: Python
- Homepage:
- Size: 1.66 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# BRICK Model Summarizer
[](https://pypi.org/project/brick-model-summarizer/)
[](https://github.com/bbartling/BrickModelSummarizer/actions)
**BRICK Model Summarizer** is a Python tool designed to validate and benchmark AI-generated BRICK models against reference models. It transforms complex BRICK schema TTL files into concise, human-readable summaries of HVAC systems, zones, meters, and central plants. By leveraging [reference BRICK models](https://brickschema.org/resources/#reference-brick-models), this tool enables users to validate AI-created models for consistency, accuracy, and adherence to expected standards.
## Purpose
The primary purpose of this repository is to provide a framework for summarizing BRICK models into HVAC-centric insights. This is especially useful for:
- **Benchmarking AI-generated BRICK models** against reference models.
- **Validating BRICK schemas** for completeness and alignment with building system expectations.
- **Empowering building engineers, analysts, and AI developers** with clear summaries of mechanical systems and operational data.
## Key Features
- **HVAC-Focused Summarization**: Extracts key details about AHUs, VAVs, meters, and central plant equipment.
- **Model Validation**: Provides a framework for benchmarking AI-created BRICK models.
- **Scalable Processing**: Processes individual or multiple BRICK schema TTL files.
## Installation
Tested in Linux and Windows Subsystem for Linux
```bash
pip install brick-model-summarizer
```
### Local Installation for development purposes
1. **Clone the repository**:
```bash
git clone https://github.com/bbartling/brick-model-summarizer.git
cd brick-model-summarizer
```
2. **Set up a virtual environment** (optional but recommended):
```bash
python -m venv env
source env/bin/activate
```
3. **Install the package locally**:
```bash
pip install .
```
---
## Usage
The package includes functions for summarizing BRICK models and generating detailed outputs. Below is an example of how to use the tool in Python to generate JSON-style data.
### Example: Processing a BRICK Model
```python
import os
from brick_model_summarizer import (
load_graph_once,
get_class_tag_summary,
get_ahu_information,
get_zone_information,
get_building_information,
get_meter_information,
get_central_plant_information,
get_vav_boxes_per_ahu,
)
# Get the absolute path of the project root
script_dir = os.path.dirname(os.path.abspath(__file__))
project_root = os.path.dirname(script_dir)
# Construct the relative path to the BRICK model
brick_model_path = os.path.join(project_root, "sample_brick_models", "diggs.ttl")
print("Resolved Brick Model Path:", brick_model_path)
# Load the RDF graph once
graph = load_graph_once(brick_model_path)
# Get the individual data components
ahu_data = get_ahu_information(graph)
print("ahu_data \n", ahu_data)
zone_info = get_zone_information(graph)
print("zone_info \n", zone_info)
class_tag_sum = get_class_tag_summary(graph)
print("class_tag_sum \n", class_tag_sum)
building_data = get_building_information(graph)
print("building_data \n", building_data)
meter_data = get_meter_information(graph)
print("meter_data \n", meter_data)
central_plant_data = get_central_plant_information(graph)
print("central_plant_data \n", central_plant_data)
vav_boxes_per_ahu = get_vav_boxes_per_ahu(graph)
print("vav_boxes_per_ahu \n", vav_boxes_per_ahu)
```
### Example Output
```python
=== AHU DEBUG Summary ===
Processed AHU's: 0
ahu_data
{'total_ahus': 0, 'constant_volume_ahus': 0, 'variable_air_volume_ahus': 0, 'ahus_with_cooling_coil': 0, 'ahus_with_heating_coil': 0, 'ahus_with_return_fans': 0, 'ahus_with_supply_fans': 0, 'ahus_with_return_air_temp_sensors': 0, 'ahus_with_mixing_air_temp_sensors': 0, 'ahus_with_supply_air_temp_sensors': 0, 'ahus_with_supply_air_temp_setpoints': 0, 'ahus_with_static_pressure_sensors': 0, 'ahus_with_static_pressure_setpoints': 0, 'ahus_with_air_flow_sensors': 0, 'ahus_with_air_flow_setpoints': 0, 'ahus_with_active_chilled_beams': 0, 'ahus_with_chilled_beams': 0, 'ahus_with_passive_chilled_beams': 0, 'ahus_with_heat_wheels': 0, 'ahus_with_heat_wheel_vfds': 0}
zone_info
{'zone_air_temperature_setpoints_found': False, 'total_variable_air_volume_boxes': 59, 'total_variable_air_volume_boxes_with_reheat': 0, 'number_of_vav_boxes_per_ahu': {}, 'vav_boxes_with_reheat_valve_command': 0, 'vav_boxes_with_air_flow_sensors': 0, 'vav_boxes_with_supply_air_temp_sensors': 0, 'vav_boxes_with_air_flow_setpoints': 0, 'co2_sensor_count': 0, 'co2_setpoint_count': 0, 'zone_air_conditioning_mode_status_count': 0, 'cooling_temp_setpoint_count': 0, 'dewpoint_sensor_count': 0, 'heating_temp_setpoint_count': 0, 'humidity_sensor_count': 0, 'humidity_setpoint_count': 0, 'temperature_sensor_count': 0, 'temperature_setpoint_count': 0, 'zone_count': 0, 'reheat_command_count': 0, 'reheat_hot_water_system_count': 0, 'reheat_valve_count': 0}
Class Similarities:
class_tag_sum
{'class_mismatches': [], 'tag_mismatches': []}
building_data
{'building_area': 'not_available', 'number_of_floors': 'not_available', 'hvac_equipment_count': 9, 'hvac_zone_count': 0}
meter_data
{'chilled_water_meter_present': False, 'hot_water_meter_present': False, 'building_electrical_meter_present': False, 'building_gas_meter_present': False, 'building_water_meter_present': False, 'electric_energy_sensor_count': 0, 'electric_power_sensor_count': 0, 'active_power_sensor_count': 0, 'ev_charging_hub_count': 0, 'ev_charging_port_count': 0, 'ev_charging_station_count': 0, 'electrical_energy_usage_sensor_count': 0, 'pv_generation_system_count': 0, 'pv_panel_count': 0, 'photovoltaic_array_count': 0, 'photovoltaic_current_output_sensor_count': 0, 'photovoltaic_inverter_count': 0, 'peak_demand_sensor_count': 0, 'people_count_sensor_count': 0}
central_plant_data
{'chiller_count': 0, 'water_cooled_chiller_count': 0, 'air_cooled_chiller_count': 0, 'centrifugal_chiller_count': 0, 'absorption_chiller_count': 0, 'boiler_count': 0, 'natural_gas_boiler_count': 0, 'noncondensing_natural_gas_boiler_count': 0, 'condensing_natural_gas_boiler_count': 0, 'electric_boiler_count': 0, 'cooling_tower_count': 0, 'cooling_tower_fan_count': 0, 'heat_exchanger_count': 0, 'heat_exchanger_discharge_temp_sensor_count': 0, 'heat_exchanger_leaving_temp_sensor_count': 0, 'heat_exchanger_supply_temp_sensor_count': 0, 'heat_exchanger_system_enable_status_count': 0, 'heat_pump_air_source_condensing_unit_count': 0, 'heat_pump_condensing_unit_count': 0, 'heat_pump_ground_source_condensing_unit_count': 0, 'heat_pump_water_source_condensing_unit_count': 0, 'heat_recovery_air_source_condensing_unit_count': 0, 'heat_recovery_condensing_unit_count': 0, 'heat_recovery_hot_water_system_count': 0, 'heat_recovery_water_source_condensing_unit_count': 0, 'hot_water_system_count': 0, 'water_pump_count': 0, 'chilled_water_system_count': 0, 'condenser_water_loop_count': 0, 'condenser_water_pump_count': 0, 'condenser_water_system_count': 0, 'domestic_hot_water_system_count': 0, 'preheat_hot_water_system_count': 0, 'radiation_hot_water_system_count': 0, 'reheat_hot_water_system_count': 0, 'water_system_count': 0, 'water_system': 1, 'water_pump': 4, 'hot_water_system': 1, 'chiller_water_flow_count': 0, 'boiler_water_flow_count': 0, 'cooling_tower_temp_count': 0}
vav_boxes_per_ahu
{}
```
One note on the output of the `Class Similarities` is it finds mismatched BRICK classes and tags by comparing them to the most current standard. If a mismatch is found, it returns a dictionary like data in the format of `('custom_tag', 'standard_tag', 0.90)`:
```python
{
'class_mismatches': [('Air_Handler_Unit', 'Air_Handling_Unit', 0.85)],
'tag_mismatches': [('custom_tag', 'standard_tag', 0.90)]
}
```
Here, **0.85 and 0.90** are similarity scores from `SequenceMatcher`, which measure how close the custom class or tag is to the standard one. These values provide a **statistical similarity percentage** from the Python `difflib` package, helping you assess how much a custom class deviates from the standard.
---
### Validating AI-Generated Models
Use the outputs to compare AI-created models against reference BRICK models, checking for consistency in:
- Equipment classification (e.g., AHUs, VAVs).
- Sensor and control points.
- Central plant configurations.
## Sample Data
Reference BRICK models from [BRICK resources](https://brickschema.org/resources/#reference-brick-models) are included in the `sample_brick_models` directory. These files can be used for testing and validation.
## Web App Demo
View a web app interface on Bens Pythonanywhere account for free!
* https://bensapi.pythonanywhere.com/
* Upload and process `.ttl` files to generate detailed BRICK model summaries.
* Compare your AI-generated models with [official BRICK Reference Models](https://brickschema.org/resources/#reference-brick-models).
* Easy-to-use web interface with support for `.ttl` file validation.
> * Also, see an example of a [client web request POST script](https://github.com/bbartling/BrickModelSummarizer/blob/develop/flask_app/client_post.py) for the PythonAnywhere web app API. A **FUTURE TODO** is to experiment with LLMs using an API like this to verify that AI-generated data models are created properly. If errors are detected, the AI should initiate a fine-tuning process for the data, especially if using an autonomous agent framework. 🚀
## Contributing
We welcome contributions to improve the repository. Please submit issues or pull requests to discuss new features, bug fixes, or enhancements.
## Roadmap
### Planned Enhancements
- **ECM and KPI Suggestions**: Develop functionality to recommend energy conservation measures (ECMs) based on model summaries.
- **Advanced Validation**: Add checks for missing or inconsistent relationships in AI-generated models.
- **PyPI Distribution**: Prepare the package for publication on PyPI.
## License
This project is licensed under the MIT License. See the LICENSE file for details.