Fusing Physics-based and Machine Learning Models for Building Energy Modeling

Buildings are major consumers of energy, primarily due to the operation of heating, ventilation, and air-conditioning (HVAC) systems. Each building is distinct with differences in building characteristics (geometry, materials, …), operating conditions (occupancy, lighting, heating, …) and boundary conditions (solar irradiance, wind velocity, …). Developing a digital model that accurately reflects these factors can help optimize the operation of energy systems. The foundation for such optimal operation decisions represents the modeling of building thermal dynamics. In this context, traditional physics-based models offer high physical interpretability, but their modeling process is complex and requires expertise. Data-driven models have also been applied with success, though they often lack interpretability or are physically inconsistent. A promising approach for thermal dynamics modeling involves combining traditional physics-based models together with data-driven models to leverage their respective advantages. The combination itself can be constructed in different ways, each exhibiting its particular characteristics. A preliminary case study used the Energyplus software as the physics-based model, combining it with different data-driven models and showed encouraging results in the context of building temperature dynamics simulation. These observations open several pathways for further research.

Buildings are major consumers of energy, primarily due to the operation of heating, ventilation, and air-conditioning (HVAC) systems. Each building is distinct with differences in building characteristics (geometry, materials, …), operating conditions (occupancy, lighting, heating, …) and boundary conditions (solar irradiance, wind velocity, …). Developing a digital model that accurately reflects these factors can help optimize the operation of energy systems. The foundation for such optimal operation decisions represents the modeling of building thermal dynamics.

In this context, traditional physics-based models offer high physical interpretability, but their modeling process is complex and requires expertise. Data-driven models have also been applied with success, though they often lack interpretability or are physically inconsistent.

A promising approach for thermal dynamics modeling involves combining traditional physics-based models together with data-driven models to leverage their respective advantages. The combination itself can be constructed in different ways, each exhibiting its particular characteristics. A preliminary case study used the Energyplus software as the physics-based model, combining it with different data-driven models and showed encouraging results in the context of building temperature dynamics simulation. These observations open several pathways for further research.

This project focuses on exploring the combination of physics-based and data-driven models for building thermal modeling. The main objective is to investigate the capabilities of such hybrid models regarding predictive accuracy, physical consistency, and data dependency. The general tasks are as follows: - Conduct a comprehensive literature review about building thermal modeling techniques and hybrid models - Explore hybrid models for time series forecasting and time series regression - Evaluate proposed methodologies using data associated with a research unit at EMPA and compare it to a variety of benchmark models. **Prerequisites** Prospective candidates are expected to possess a foundational understanding of statistics, optimization, machine learning, and practical experience in programming in Python. Knowledge of HVAC systems and building energy modeling is a plus. Interested students are asked to submit an updated CV along with transcripts of academic records.

This project focuses on exploring the combination of physics-based and data-driven models for building thermal modeling. The main objective is to investigate the capabilities of such hybrid models regarding predictive accuracy, physical consistency, and data dependency. The general tasks are as follows:

- Conduct a comprehensive literature review about building thermal modeling techniques and hybrid models

- Explore hybrid models for time series forecasting and time series regression

- Evaluate proposed methodologies using data associated with a research unit at EMPA and compare it to a variety of benchmark models.

**Prerequisites**

Prospective candidates are expected to possess a foundational understanding of statistics, optimization, machine learning, and practical experience in programming in Python. Knowledge of HVAC systems and building energy modeling is a plus. Interested students are asked to submit an updated CV along with transcripts of academic records.

leandro.vonkrannichfeldt@epfl.ch

leandro.vonkrannichfeldt@epfl.ch