HVAC: Specialized component library for Heating, Ventilation, Air Conditioning and Refrigeration

JuliaSim HVAC provides a library with pre-built components for acausal modeling and dynamic simulation of HVAC systems. Physics-based models for multi-phase heat exchangers, compressors, valves etc. are developed using ModelingToolkit and integrated with refrigerant thermodynamic property models that leverage fast and accurate spline interpolations. The library seamlessly integrates with the rest of the JuliaSim Scientific Machine Learning ecosystem for automated model calibration to plant data, surrogate modeling for accelerated simulation, discovery of unknown physics and model-based control.

Overview:

HVAC systems involve physical phenomena such as two-phase compressible, viscous and turbulent thermal-fluid flow. The resulting system has stiff non-linear dynamics (in that pressure dynamics change in the order of <1 second while thermal dynamics of the conditioned space change in the order of minutes) in addition to discontinuous changes in thermodynamic properties close to phase change boundaries. Current design workflows use disparate tools for each step: modeling & simulation, model calibration, design optimization, machine learning, control and deployment with information (such as derivatives, solver configurations) lost in each switch.

JuliaSim HVAC provides a comprehensive suite for modeling and simulation that integrates with tailored stiffness and discontinuity-aware numerical solvers all of which are composable with scientific machine learning workflows to enable design and operation of the next generation of HVAC systems.

Features:

1. Model library:

• Configurable Tube-Fin Heat Exchangers

• Compressors: TypicalCompressor

• Valves: TypicalLEV

• Fans: SimpleFan

• Configurable conditioned spaces

• Pipes

2. Thermodynamic Property Models:

• Spline-based thermodynamic property models for several refrigerants (such as R32, R1234YF, R290, R152a, R134a, R410A, R717).
• Dry Air
• Moist Air

detailed in BaseProperties

3. Scientific Machine Learning Workflows:

• JuliaSim Model Optimizer Automate model calibration unleashing the power of automatic differentiation of the simulator. Leverage machine learning to discover new relationships (such as compressor performance or pressure drop correlations).
• JuliaSim Digital Echo Accelerated HVAC simulation using neural surrogates
• JuliaSim Control Comprehensive tools for Feedback and Model Predictive Control (MPC)

Getting Started:

• Build a Vapor Compression Cycle: Simple Refrigerant Circuit
• Model a double-sided air and refrigerant tube-fin heat exchanger with a cross-flow configuration: Tube-Fin tutorial