Jessica Hernandez

[jhernandez312]

Weekly Notebook Entry — Week [01]

Overview

• Week Span: [9/9] to [9/15]

Tasks for This Week

List your specific objectives for this week, including learning targets and project milestones.

• [ ] Task 1: Understand 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: [Task 2] - (completed)

• Description: Looked at DOE reference buildings/ComStock and if we could use this for the multivariate regression model
• Link to Work:
• DOE reference buildings
• Comstock
• Outcome & Reflection:
• DOE reference buildings contain 16 different variants of commercial buildings across 16 different regions (one of which is Atlanta). Each of the 16 buildings are quite detailed and idf files are included.
• Comstock contains information about multiple buildings. Each was a result of energy simulations that took in a mix of the DOE reference buildings + actual data.

Task: [Task 1] - (ongoing)

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

Challenges & Solutions

• Figuring out if we need a new model? and what the multivariate model takes as input/can we somehow use the comstock data

General Takeaway, Skills & Knowledge Acquired

• IDF files may be useful to reconstruct the DOE buildings so we can obtain the information we're missing

Team Meetings

• Meeting Highlights:
• Parametrics models can provide another way to obtain data
• Our scope can be the GT campus
• UMI is a model that's able to simulate energy. Perhaps we can use this?
• How are we going to build an architectural model with the random data that we have so that it can predict the heating and cooling load?
• Does energy demand and energy consumption mean the same thing for the 2 models?
• Peer Feedback:

Personal Reflection

• There's quite a bit of tools that exist to simulate energy consumption (whether that be physics-based, or other). While we're not using the demand ninja model, the paper provides insight on the various attempts in the during that time and is a nice summary.

Visual Documentation (optional)

Additional Notes & Resources (optional)

• Resource Name - Brief description.

---

Plans for Next Week

• [ ] Next Week's Task 1: Figure out what the models need as inputs

[jhernandez312]

Weekly Notebook Entry — Week [02]

Overview

• Week Span: [9/16] to [9/22]

Tasks for This Week

List your specific objectives for this week, including learning targets and project milestones.

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

Tasks

Task: [Task 1] - (completed)

• Description:
• Link to Work:

• Outcome & Reflection:
• We might be able to recreate the inputs required for the multivariate regression model (specifically, compactness) by some combination of the already available data

General Takeaway, Skills & Knowledge Acquired

• The residential group were also facing similar issues. From the general meetings on Mondays, the microclimate team was also facing a similar issue of narrowing down the inputs to a feasible amount (from 55 down to 12ish).

Team Meetings

• Meeting Highlights:

  Title of our project - (similar to Pix2daylight)

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

Users of the project (Who can benefit from it?/Who are we creating the surrogate model for?

1. Potential users - architects / governments - quick simulations to test effects at scale

Problem Definition

1. Lack of easy way to model energy performance for large number of buildings - commercial. Generate data from white box simulations and then use for a black box approach. Grey box approach - Combination of black and white box
2. Single building vs Urban Scale
3. Neighborhoods Umi - example of neighborhood scale tool

• Peer Feedback:

Personal Reflection

• In addition to different energy models predicting through different methodologies (physics-based, RL, etc.), models can differ through scale. This is highlighted through our team meetings since the architects need to build models at either a building, neighborhood, or city scale.

Visual Documentation (optional)

Additional Notes & Resources (optional)

• Resource Name - Brief description.

---

Plans for Next Week

• [ ] Next Week's Task 1: Look at the Google dataset/model has context menu

[jhernandez312]

Weekly Notebook Entry — Week [03]

Overview

• Week Span: [9/23] to [9/29]

Tasks for This Week

List your specific objectives for this week, including learning targets and project milestones.

• [ ] Task 1: Set up Google TensorFlow Smart Buildings Simulator

Tasks

Task: [Task 1] - (completed)

• Description:

• set up Google TensorFlow Smart Buildings Simulator

• Link to Work:

• 

• Outcome & Reflection:

• Took longer than expected to set up because there were issues with the virtual machine. It was running incredibly slow and the fix was to disable Windows' Optional Features. Had to go to System32 folder -> OptionalFeatures.exe -> and disable "Windows Subsystem for Linux", "Windows Hypervisor Platform" and "Virtual Machine Platform". The one that actually made a difference was disabling Windows Subsystem for Linux, but I didn't want to try all combinations. Also had to go into Windows Security -> Memory Integrity. I'm using the Oracle VirtualBox for the virtual machine instead of VMWare since Ubuntu was crashing after being installed in VMWare. Perhaps it works now that Windows Hypervisor Platform is off, but VirtualBox works fine. Another option was installing Ubuntu on my laptop, but in theory that seemed like the long options.

• As for the Google Dataset & RL model, the model has a couple of stages. Given the floorplans & metadata of HVAC systems of a buildings, it's able to simulate heat/energy consumption in a physics-based way. You can then optimize it by giving it real info?

• Task: [Google Model] - (ongoing)

• Current Status:

• in progress

• Expected Completion:

  • next week

    Challenges & Solutions

• Understand google model through either running it or reading the paper (prob a mix of both)

General Takeaway, Skills & Knowledge Acquired

• Ubuntu uses the command line interface for lots of things including program installation. For github tools, the common option is to clone entire github repositories. For python libraries or even VSCode, you can type a command (i.e. pip install or sudo apt install __)
• Ubuntu's keyboard shortcuts are similar to Windows'.

Team Meetings

• Meeting Highlights:

• 

• Peer Feedback:

Personal Reflection

• Ubuntu is fun to use
• Google's model seems to be for a single building at a time (however, it can adapt to a lot of different commercial buildings)

Visual Documentation (optional)

Additional Notes & Resources (optional)

• Resource Name - Brief description.

---

Plans for Next Week

• [ ] Next Week's Task 1: Investigate the simulator's scalability and how it can be applied to multi-building simulations for more comprehensive energy management solutions.

[jhernandez312]

Weekly Notebook Entry — Week [04]

Overview

• Week Span: [9/30] to [10/6]

Tasks for This Week

List your specific objectives for this week, including learning targets and project milestones.

• [ ] Task 1: Investigate the simulator's scalability and how it can be applied to multi-building simulations for more comprehensive energy management solutions.

Task: [Task 1] - (ongoing)

• Current Status:
• in progress
• Expected Completion:
• Monday

Challenges & Solutions

-challenges:

General Takeaway, Skills & Knowledge Acquired

-

Task: [Task 2] - (ongoing)

• Current Status:
• in progress
• Expected Completion:
• Monday

Team Meetings

• Meeting Highlights:
• Data to Geometry Workflow - Hang showed how to transform comstock idf files to retrieve geometry (window to wall ratio, area, etc). Comstock contains lots of osm files and it would be impossible to go through it file by file. One challenge would be to figure out a workflow
• Next steps are how to automate the process of converting large data rows efficiently
• Peer Feedback:

Personal Reflection

• Features are able to be reconstructed which means we can use the comstock data into the multivariate regression model.

Plans for Next Week

• [ ] Next Week's Task 1: Run multivariate regression model has context menu