Anubha Mahajan

[amahajan68]

Weekly Notebook Entry — Week 4

Overview

• Week Span: 9/9 to 9/15

Tasks for This Week

• [ ] Task 1: Give an overview of the models that were previously used (Multivariate Regression, Demand Ninja)
• [ ] Task 2: Figure out the important commercial building features-- ComStock and DOE reference buildings

Tasks

Task: Investigate Commercial Building Features - (completed)

• Description: Looked at DOE reference buildings and ComStock to see if we could use them for the multivariate regression model.
• Link to Work:
  • DOE reference buildings
  • Comstock
• Outcome & Reflection:
  • The DOE reference buildings cover 16 different types of commercial buildings across various U.S. regions, including Atlanta. These models are highly detailed, and their IDF files can be used in our regression model.
  • ComStock contains multiple building simulations that combine DOE reference buildings with real-world data. This could be a useful resource for expanding our dataset and improving model accuracy.

Task: Overview of Previously Used Models - (ongoing)

• Current Status: In progress
• Expected Completion: Next week

Task: Create a Pipeline for Multivariate Regression Model - (completed)

• Description: I used the Mockaroo website to generate mock data, which was then input into the Multivariate Regression Model. This mock data simulated different building parameters like glazing area, surface area, wall area, relative compactness, and more. The purpose was to ensure we have a fully functioning pipeline for energy consumption predictions based on these parameters.

• Link to Work:

  • Screenshot of the pipeline:

• Outcome & Reflection:

  • By using Mockaroo, we successfully generated realistic data to test the Multivariate Regression Model. The pipeline is now capable of accepting various inputs and generating predictions for heating and cooling loads based on those parameters. This process was essential for ensuring that the model is flexible and can handle different building configurations for energy simulation.

Challenges & Solutions

• Challenge: Determining whether we need a new model and understanding the inputs for the multivariate regression model. The question of whether we can use ComStock data for the regression model remains.

  • Solution: Further research is needed on how ComStock data can complement or replace certain inputs in the multivariate model.

• Challenge: Generating realistic input data for the multivariate regression model.

  • Solution: Used Mockaroo to generate structured mock data for key building characteristics such as glazing area, wall area, and relative compactness. These were then used as inputs to the regression model.

General Takeaway, Skills & Knowledge Acquired

• Takeaway: IDF files from DOE reference buildings might allow us to reconstruct missing data for the regression model and help simulate buildings for Atlanta-specific energy predictions. Additionally, using Mockaroo as a tool for generating mock data ensures flexibility and scalability in testing the model.

• Skills Acquired:

  • Generating structured mock data using Mockaroo.
  • Setting up a data pipeline for multivariate regression modeling.
  • Understanding commercial building features from DOE reference models and ComStock.

Team Meetings

• Meeting Highlights:

  • Discussed the use of parametric models to generate data for the energy consumption predictions.
  • Explored the possibility of narrowing our scope to GT Campus for model simulation.
  • Reviewed tools like UMI, which can simulate energy at the urban scale. This might be helpful for large-scale energy predictions.
  • Key Questions:
  • How do we build an architectural model that predicts heating and cooling loads with random data?
  • Does "energy demand" mean the same thing as "energy consumption" across different models?

• Peer Feedback: Focus more on automating parametric models and integrating them into the regression pipeline.

Personal Reflection

• There are numerous tools available for simulating energy consumption, including both physics-based and data-driven methods. While our project won’t directly use the Demand Ninja model, reviewing its approach gave us helpful insights into how energy models have evolved over time.
• Setting up the pipeline with Mockaroo allowed us to easily experiment with different building configurations, which will help refine the model's performance moving forward.

Additional Notes & Resources

• UMI - Link - An urban energy modeling tool that might be useful for simulating large-scale energy consumption.

---

Plans for Next Week

• [ ] Next Week's Task 1: Figure out what the models need as inputs for more accurate predictions.

[amahajan68]

Weekly Notebook Entry — Week 5

Overview

• Week Span: 9/16 to 9/22

Tasks for This Week

• [ ] Task 1: Figure out what the models need as inputs

Tasks

Task: Figure Out Model Inputs - (completed)

• Description: Investigated inputs required for the multivariate regression model, specifically focusing on compactness and other key features, to determine if available data can be used to recreate these inputs for energy prediction.

• Link to Work:

  

• Outcome & Reflection: We identified that inputs such as compactness can potentially be recreated using existing data, which simplifies the process of input preparation for the regression model. This realization also opens the door for using synthetic datasets based on building geometry and footprint.

General Takeaway, Skills & Knowledge Acquired

• Gained insight into the process of refining model inputs, a challenge that other teams (microclimate, residential) also faced during their model development. Narrowing down inputs while maintaining model accuracy is crucial.

Team Meetings

• Meeting Highlights:

  • Discussed project titles:

    1. UEP - Urban Energy Predictor
    2. REAP - Regression Energy Analytics and Prediction

  • Identified potential users of the surrogate model:

  • Architects and government planners who require quick simulations to assess the energy impact of design changes at scale.

  • Discussed the problem definition:

  • The challenge of modeling energy performance for a large number of buildings at an urban scale.

  • Emphasized the need to combine white-box simulations (detailed, physics-based) with black-box models (machine learning) to create a scalable, grey-box solution.

  • Explored neighborhood-scale tools such as UMI for integrating energy performance across multiple buildings.

  • TODOs from the meeting:

  • Explore UMI and request Georgia Tech building data (Joseph).

  • Build a 3D model for a specific Atlanta neighborhood.

  • Use OpenStreetMap for building footprint and height data.

  • Investigate the Google Dataset for energy performance (Anubha).

  • Look into tools that can extract window-to-wall ratio or other building geometry details.

• Peer Feedback:

  • Focus on refining input data and investigating neighborhood-scale simulation tools.

Personal Reflection

• The discussion on scaling models from single buildings to urban neighborhoods was insightful, reinforcing the importance of handling different scales in architectural and energy modeling. The possibility of using tools like UMI and OpenStreetMap for neighborhood-scale modeling excites me as we move forward with integrating different data sources.

Visual Documentation (optional)

Additional Notes & Resources (optional)

• UMI Tool - Link - An example of a neighborhood-scale energy modeling tool.

---

Plans for Next Week

• [ ] Next Week's Task 1: Look at the Google dataset/model and begin integrating it with neighborhood-scale data for energy predictions.

[amahajan68]

Weekly Notebook Entry — Week 6

Overview

• Week Span: 9/23 to 9/29

Tasks for This Week

• [ ] Task 1: Reviewed Google Paper and SMART Dataset
• [ ] Task 2: Set up a virtual environment to simulate a Linux system for running the model

Key Insights from "A Lightweight Calibrated Simulation Enabling Efficient Offline Learning for Optimal Control of Real Buildings"

1. Customized Simulation Approach: The paper presents a lightweight simulator customized for each building by calibrating it to telemetry data. The simulator achieved only 0.64°C drift over six hours, making it highly accurate for real-world use.

2. RL for HVAC Control: The RL agent optimizes HVAC energy consumption by adjusting setpoints to balance energy use, carbon emissions, and occupant comfort.

3. Simulator Configuration & Scalability: The simulator is easily configurable for different buildings using floorplans and HVAC metadata. While currently designed for single buildings, the paper highlights potential scalability to city-scale energy management.

4. Calibration Method: Finite Differences (FD) is used for modeling thermal diffusion, ensuring accurate simulation of heat transfer across the building. Real-world telemetry data is used to calibrate the simulator.

5. Performance: The simulator demonstrated Mean Absolute Error (MAE) as low as 0.64°C, validating its accuracy across different time intervals and conditions.

6. Future Work: The authors plan to improve generalization by tuning across multiple seasons and expanding the action space to include additional HVAC parameters.

7. Potential for Expansion: The approach can be adapted for multi-building environments, offering the potential for city-wide energy optimization.

Tasks

Task: Review Google Paper & SMART Buildings Dataset - (completed)

• Description: I reviewed Google's TensorFlow Smart Buildings Simulator and the accompanying SMART Buildings Dataset, which includes six years of telemetry data from three commercial office buildings. I also examined the paper from BuildSys '23, "A Lightweight Calibrated Simulation Enabling Efficient Offline Learning for Optimal Control of Real Buildings."

• Link to Work: Google Paper on Smart Buildings

• Outcome & Reflection: The paper provided key insights into RL-based HVAC optimization and calibration techniques. The simulator achieves less than 0.64°C drift over six hours, making it highly effective for single-building energy management. There is potential for adapting the model to simulate multiple buildings for larger-scale energy optimization solutions.

Task: Set Up Virtual Environment for TensorFlow Smart Buildings Simulator (completed)

• Description: Set up a Linux environment using Oracle VirtualBox to run the Google TensorFlow Smart Buildings Simulator. This provided the necessary dependencies for running the model in a Linux-based environment.

• Outcome & Reflection: After troubleshooting performance issues related to the virtual machine, switching to Oracle VirtualBox resolved the issues. The Linux environment is now fully operational for running the simulator and making further modifications.

Task: Investigate Reinforcement Learning Techniques for Energy Optimization (ongoing)

• Current Status: Currently researching the RL models used in the Google simulator and focusing on how they can be scaled to optimize energy consumption across multiple buildings.

• Expected Completion: Next week

Challenges & Solutions

• Challenge: Understanding the Finite Differences approximation used in the calibration process.
  • Solution: Continued research into simulation fidelity techniques and exploring how to integrate additional telemetry data to expand the model for multi-building simulations.

General Takeaway, Skills & Knowledge Acquired

• Takeaway: Combining physics-based simulations with real-world data calibration is highly effective for RL training in HVAC systems. Expanding the simulator for multi-building optimization could unlock new possibilities in urban energy management.

• Skills Acquired:

  • Reinforcement Learning (RL) for energy optimization
  • Finite Differences method for thermal modeling
  • Calibration using real-world data
  • Setting up and troubleshooting a Linux virtual environment for model simulation

Personal Reflection

• This week reinforced my interest in using RL for energy optimization in real-world settings. Expanding the simulator from individual buildings to multi-building scenarios offers exciting possibilities for future projects in urban sustainability.

Visual Documentation (optional)

Additional Notes & Resources (optional)

• TensorFlow Smart Buildings Simulator - GitHub Repository - Open-source tool for training RL agents to optimize energy usage in buildings.

---

Plans for Next Week

• [ ] Next Week's Task 1: Investigate how the simulator's calibration techniques can be scaled for multi-building simulations.

[amahajan68]

Weekly Notebook Entry — Week 7

Overview

• Week Span: 9/30 to 10/6

Tasks for This Week

• [ ] Task 1: Explore and automate the Data to Geometry workflow for extracting Comstock data for multivariate regression models.

Tasks

Task: Data to Geometry Workflow - (ongoing)

• Description: Explored the process of transforming Comstock IDF files to extract geometric features such as window-to-wall ratio, area, and other building attributes. The goal is to automate this workflow and integrate it with our multivariate regression model for energy consumption analysis.

• Current Status: The initial steps for extracting geometric data from Comstock IDF files were completed. We are now working on automating the workflow for large datasets.

• Expected Completion: Monday

Challenges & Solutions

• Challenge: Automating the extraction of geometric features from a large number of Comstock IDF files.
  • Solution: We started implementing batch processing to handle large datasets more efficiently and to avoid manual extraction on a file-by-file basis.

General Takeaway, Skills & Knowledge Acquired

• Learned how to extract relevant geometric features from Comstock IDF files.
• Gained experience in batch processing and automating workflows for large datasets.

Team Meetings

• Meeting Highlights:
  • Discussed the best approach for automating the data extraction process from Comstock IDF files.
  • Identified potential bottlenecks in file processing and began addressing them.

Personal Reflection

• The work this week on automating the Data to Geometry workflow has demonstrated the importance of efficiency when working with large datasets. Automating this process will not only save time but also improve the accuracy of our multivariate regression model.

Plans for Next Week

• [ ] Next Week's Task 1: Complete automation of the Data to Geometry workflow and integrate extracted features into the multivariate regression model.