According to the U.S. DOE Building Energy Data Book (2010), the building sector is responsible for 40% of the nation’s primary energy consumption (Gelman 2011). Historically, the building envelope has been tasked with the challenge of neutralizing highly variable solar, air, pollutant, and moisture conditions, all while maintaining a constant desired interior condition (Selkowitz 2003). In an attempt to maintain indoor conditions, modern building engineering rely heavily on fossil-fueled mechanisms and highly insulated or glazed static building envelopes to reject exterior energy flows (Krarti 2010). To achieve significant progress towards global targets for clean on-site energy self-sufficiency within the building sector, integrating building envelopes with multifunctional energy transformation systems could provided a series of benefits such as: electrical generation, hydrothermal collection, daylighting, reduced cooling loads, humid air dehumidification, water recuperation, distributed heating and cooling, and improved human comfort and well being.
In parallel, current energy modeling tools lack the fidelity and adaptability necessary to validate the multi-functional benefits of next-generation envelope technologies (Wetter 2011). For example, EnergyPlus, a common practice building energy modeling (BEM) tool (Crawley 2001), has been designed to assess energy performance of a wide range of conventional building systems and whole buildings but produce significant inaccuracies in the thermal and airflow behavior of a double skin facade when compared to measurements (Kim 2011). In response, high fidelity simulations, such as computational fluid dynamics (CFD) (Guide 2013) and raytracing (Ward 2011), or device specific models, following the known physics of the system, are created in isolation which can accurately model the mass, momentum, and energy transfer of complex building systems. Compuational fluid dynamics has been used in architectural and engineering studies of double-skin facades for a number of years. CFD was used to analyze turbulent natural ventilation in a double-skin facade with thermal mass louvres and found reasonable convergence and validation with experimental data (El-Sadi 2010). Due to the complexity of computational fluid dynamics, simulation are often difficult to adapt to new configurations and not readily integrated with common practice building energy models (Hensen 2002).
In practice, these models lead to accurate results but require tremendous expertise to create and poorly adapt to the rapid iteration of architecture practice. These models are often created separately from building energy models and loosely connected through pre- or post-processing with building energy data to assess system performance. In post-processing, each model is run separately and results are compiled in a spreadsheet application. Although post-processing is widely used in practice, the organization of data is time consuming for the user and it is the least accurate method because there is no feedback between the system model and building energy model. Pre-processing fully simulates one model and uses the results as input to the second model; commonly used when modeling dynamic glazing systems for improved natural daylight with adaptive electric lighting building energy models. Pre-processing is an improvement over post-processing as it provides better feedback between the models but often only one directional. However, pre-processing requires the expertise need to simulate the primary model, reformat the results as input, and setup the BEM to utilize the input; creating a workflow that is difficult for laypersons to use and reducing the ability to iterate quickly through different designs. Development of active systems is impeded by current modeling workflows which do not provide adequate feedback or facilitate rapid design iteration within the context of build energy modeling (BEM) (Hensen 2012). To integrate and characterize emerging climate responsive technologies, an approach to modeling is required that encourages information exchange between different types of models at different scales, such that adaptive, higher-fidelity models can interface with standard BEM frameworks.
From current research and projections, future simulations will need to possess a level of abstraction able to adapt and easily model new built environment technologies but accurate enough to have confident results. Device and system models are often created in isolation from the building context and loosely coupled to building energy models (BEM). The Functional Mock-Up Interface (FMI) Standard provides a framework for coupling models that streamlines the connection and exchange of data between device and system models with BEM, also known as co-simulation . 1 Current research at LBNL’s Simulation Research Group is leveraging Modelica, a language for modeling complex, dynamic systems, and functional mock-up units (FMUs), a standard for packaging models, for co-simulation of building systems with common practice BEM software, EnergyPlus (Nouidui 2013). This method of creating complex models and linking to common practice BEM software has exciting but not yet explored potential in the modeling and validation of next-generation dynamic envelopes. Adaptive, predictive, uncertainty quantified models of dynamic building envelope technologies will be developed in Modelica and Python and connected through the FMI standard with common practice BEM software, specifically EnergyPlus (and possibly Grasshopper).
A large portion of my contribution to the research will be co-simulation of the network model with EnergyPlus through the FMI standard and improving that interface for ease of use. To make the greatest impact on building energy modeling practices the model must be easy to use and provide results which answers questions relevant to architecture and engineering projects. Since EnergyPlus is a widely used building energy modeling tool by practitioners, co-simulation will occur through EnergyPlus’s ExternalInterface:FunctionalMockupUnitImport module with EnergyPlus being the master simulation program and an FMU being the slave. The FMU run by EnergyPlus will be a proxy created with C code that externally calls and interfaces with the network model written in python. At each timestep, EnergyPlus calls the FMU and waits for a response before continuing to the next building model timestep. The FMU proxy will externally call and execute the network model with inputs from the EnergyPlus co-simulation. Once the network model has completed simulation, the results are passed back to the FMU proxy which will co-simulate with EnergyPlus and progress to the next timesetp. This will allow the network model written in python to interface with the FMI standard and EnergyPlus.
“FMI for Model Exchange and Co-Simulation.” FMI-Standard.org. Modelica Association, 25 July 2014. Web.↩