BeSim Toolbox

Matlab toolbox for quick design and simulation of advanced building climate control algorithms.

Features

• Interface for linearized white-box building envelope models from Modelica
• Automated construction of model predictive control (MPC) and state estimation algorithms
• Closed-loop simulation, plotting, and performance analysis
• Approximate MPC via machine learning (deep lerarning in particular)
• For quick start and more details check out the presentation about algorithms and tools behind BeSim

Installation

tbxmanager

• Install tbxmanager
• Install BeSim via: tbxmanager install besim
• Check for updates: tbxmanager update

  manual

• clone BeSim repository
• save BeSim folder with its subfolders to Matlab path

Prerequisities

• Matlab: developed and tested on R2017a and R2017b
• Yalmip mathematical modeling and optimization toolbox (BeSim’s backbone)
• Optimization solver, e.g. Quadprog or commercial solver such as Gurobi (solution of implicit MPC and MHE problems)
• Matlab toolboxes: Deep Learning, Machine learning (approximate MPC functionality)

Getting Started - Demos

run following scripts in Matlab to get quick results:

• BeInit.m: design and simulation of optimization-based MPC and state estimator for selected building model
• BeInitML.m: design and simulation of approximate MPC via machine learning for selected buiding model

Structure

Functional Structure: Graphical overview of BuiSim structure with data-flow dependencies.

Repository Structure: List of repository folders with associated functionality.

• Be_Run (run files and demos)
• Be_Modeling (model loading, discretization, model order reduction)
• Be_Disturbances (disturbance trajectories loading)
• Be_References (reference trajectories loading)
• Be_Estimation (estimator design)
• Be_Control (controller design)
• Be_Simulation (main simulation and plotting functions)
• Be_Learn (machine learning functions, synthesis of approximate MPC)
• buildings (building models files)
• Data (stored results)

Algorithms

List of key enabling algorithms implemented in BeSim.

Model Order Reduction

• Balanced truncation

State Estimation

• Kalman filters (KF)
• Moving horizon estimation (MHE)

Optimal Control

• Model predictive control (MPC)
• Approximate MPC via machine learning

Machine Learning Models

• Deep learning (DL)
• Regression trees (RT)

Building Models

Model Structure

• Linearized white-box building envelope models
• arbitrary linear time invariant model in state space format

Available Building Models

Author

Email: jan.drgona@pnnl.gov

Ján Drgoňa
postdoctoral researcher
Pacific Northwest National Laboratory
Optimization and Control Group
Richland, WA, USA

Acknowledgement

The first stage of the toolbox emerged from the code development of the author during his PhD study held at Institute of Information Engineering, Automation, and Mathematics, Slovak University of Technology in Bratislava under the supervision of prof. Michal Kvasnica.

The second stage with detailed white-box building models was developed during the visiting PhD and post-doc position at Thermal Systems Simulation (The SySi) research group, Department of Mechanical Engineering Division of Applied Mechanics and Energy Conversion (TME), KU Leuven under the supervision of prof. Lieve Helsen.

An early contribution of Damien Picard towards the code development and building modeling, conceptual contributions of Martin Klaučo and Michal Kvasnica, and modeling work of Filip Jorissen and Iago Cupeiro Figueroa on Infrax building and borehole models are gratefully acknowledged.

The financial support by the European Union through the EU-H2020-GEOTeCH project ‘Geothermal Technology for conomic Cooling and Heating’ is acknowledged.

Finally, later stages of this work partially emerged from the IBPSA Project 1, an international project conducted under the umbrella of the International Building Performance Simulation Association (IBPSA). Project 1 will develop and demonstrate a BIM/GIS and Modelica Framework for building and community energy system design and operation.

References

1. Cupeiro Figueroa I., Drgona J., Helsen L., State estimators applied to a linear white-box geothermal borefield controller model, 16th International Building Performance Simulation Association Conference, Rome, Italy, 02 Sep 2019 - 04 Sep 2019. N/A. Proceedings of the 16th IBPSA Conference 2019

2. Cupeiro Figueroa I., Drgona J., Abdollahpouri M., Picard D., and Helsen L., State Observer for Optimal Control using White-box Building Models, Purdue Conferences - 5th International High Performance Building Conference, Purdue University, West Lafayette, IN, USA. INTERNATIONAL HIGH PERFORMANCE BUILDINGS CONFERENCE. Purdue e-Pubs. Jul 2018.

3. Drgona J., Picard D., Kvasnica M., Helsen L. (2018). Approximate model predictive building control via machine learning. APPLIED ENERGY, 218, 199-216. doi: 10.1016/j.apenergy.2018.02.156.

4. Picard D., Drgoňa J., Kvasnica M., Helsen L. (2017). Impact of the controller model complexity on model predictive control performance for buildings. Energy and Buildings, 152, 739-751.

5. Drgoňa J., Picard D., Helsen L. (2020). Cloud-based implementation of white-box model predictive control for a GEOTABS office building: A field test demonstration. Journal of Process Control, 88, 63-77.