Predict HVAC Savings With Thermal Mass

ISEF Category: Energy: Sustainable Materials and Design

Subcategory: Thermal Generation and Design  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A heavy wall can act like a battery for heat. If you place thermal mass in the right spot, it can soak up heat during the day and give it back later when the air cools. That can change how hard your HVAC system works. You can test that idea with simulation, data, and machine learning.

What Is It?

Thermal mass means a material can store heat. Concrete, brick, stone, and some phase-change materials do this well. Think of them like a sponge, but for heat instead of water. A thick stone floor can absorb warmth when the house gets hot, then release that heat later when the air temperature drops.

This project asks a simple question with a lot of moving parts: where should thermal mass go inside a house to lower heating and cooling demand? You are not just guessing. You can build home models in EnergyPlus, a building energy simulation tool, then compare how different placements change HVAC load. Machine learning can help you predict the best layout from features like room position, material type, climate, and window pattern.

Why This Is a Good Topic

This is a strong science fair topic because you can test many design choices without building a full house. You can change one placement variable at a time, measure the effect on energy use, and compare climates or room layouts. The project connects to real problems like cutting energy bills and reducing grid demand. You can also learn simulation, feature selection, model training, and how to judge whether a prediction model really works.

Research Questions

• How does thermal-mass placement in a house model change predicted HVAC load reduction?
• What is the effect of room location on the cooling benefit of passive thermal mass?
• Does adding thermal mass near sunlit windows reduce peak HVAC demand more than placing it in interior walls?
• To what extent does climate zone change the best placement strategy for thermal mass?
• Which material type, such as concrete, brick, or stone, gives the largest modeled load reduction?
• How does insulation level change the predicted benefit of thermal mass placement?

Basic Materials

• Computer with enough memory to run EnergyPlus simulations.
• EnergyPlus building simulation software.
• Spreadsheet software for organizing simulation outputs.
• Python installed with pandas, scikit-learn, and matplotlib.
• Virtual building templates or simple house geometry files.
• Public weather files for several climate zones.
• Digital notebook or lab notebook for tracking model settings.

Advanced Materials

• Access to a workstation or lab computer with strong processing power.
• EnergyPlus and OpenStudio for building model editing and simulation.
• Python with scikit-learn, numpy, pandas, and matplotlib for model building.
• Jupyter Notebook for reproducible analysis.
• Git for version control.
• ImageJ or similar image analysis software if you compare geometry layouts visually.
• Access to university HVAC or building science datasets, if available.
• Citation manager for tracking sources and simulation assumptions.

Software & Tools

• EnergyPlus: Simulates building energy use and HVAC loads under different design choices.
• OpenStudio: Helps you build and edit building models for EnergyPlus.
• Python: Runs data cleaning, model training, and error analysis.
• Jupyter Notebook: Keeps your code, plots, and notes in one place.
• scikit-learn: Builds and tests machine-learning prediction models.

Experiment Steps

1. Define the outcome you will predict, such as annual HVAC load or peak cooling demand.
2. Choose the thermal-mass variables you will change, such as placement, material type, or room location.
3. Build a small set of EnergyPlus models that differ only in those variables.
4. Run the simulations and organize the outputs into a clean dataset.
5. Train a prediction model and check whether it performs better than a simple baseline.
6. Compare feature importance or sensitivity results to explain which design choices matter most.

Common Pitfalls

• Changing more than one building feature at once, which makes it impossible to tell whether thermal mass placement caused the result.
• Using too few simulated house designs, which gives the machine-learning model weak training data.
• Comparing outputs from different weather files without controlling for climate, which confuses placement effects with weather effects.
• Treating energy savings as one number without checking peak load, seasonal load, and model error separately.
• Forgetting to test a simple baseline model, which can make a weak prediction model look better than it really is.

What Makes This Competitive

A competitive version of this project goes beyond one house model and one weather file. You can compare multiple climates, multiple room layouts, and more than one type of thermal mass. Strong projects also report model error, baseline performance, and feature importance, so the reader can see why the model works. If you add a careful sensitivity analysis, your project starts to look like real building science research.

Project Variations

• Test phase-change materials instead of concrete or brick and compare predicted load reduction.
• Compare thermal-mass placement in a single-family home, an apartment, and a tiny house model.
• Add window shading or night ventilation and measure whether those features change the best thermal-mass placement.

Learn More

• EnergyPlus Documentation: Official building simulation guides and examples from the U.S. Department of Energy, searchable online.
• OpenStudio API Documentation: Free documentation for building and EnergyPlus model workflows, searchable online.
• U.S. Department of Energy Building America: Research reports on home energy performance, available through the DOE site.
• NREL Publications Database: Free papers on building energy modeling and thermal storage, searchable on the National Renewable Energy Laboratory site.
• MIT OpenCourseWare: Free course materials in energy and building systems, searchable by course name on the MIT OpenCourseWare site.
• Google Scholar: Search for review articles on thermal mass, HVAC load, and building energy prediction.