Computational Robustness of Building Energy Models (Case: EnergyPlus)
Building energy models developed with EnergyPlus are abstracted, virtual and mathematical representations of a set of complex thermo-physical interactions of multiple energy and mass flows that originate as a result transient tensions between dynamic ambient climatic disturbances and a range of human (visual-thermal) comfort conditions required for indoor spaces. Like all sorts of models, BEMs are inherently flawed to a certain degree. This is basically due to “idealistic” representations of the use, state and behavior of the modeled object (in this case: buildings). EnergyPlus-based BEMs assume complete and well-defined parameter inputs for building and surrounding geometry, envelope construction assemblies, internal gains (from occupants, lights, and appliances) and dynamic operational and occupancy schedules. BEMs take all of these inputs with 100% certainty that they perfectly represent the true properties of the building that is being modeled. BEMs are developed to operate with deterministically assumed input parameters. However, as most BEM experts state, actual building properties and its usage behavior comes with a considerable amount of uncertainty the magnitude of which is not typically taken into account in today’s BEM tools. Therefore, the disagreement between simulated and measured errors should not be perceived due to a lack of sophistication in the predictive capacity of mathematical models of BEM tools but due to BEM tools missing capability to represent that actual building behavior with all the inherent uncertainties in the input parameters. The physics behind EnergyPlus is well-developed and sophisticated and BEM programs like EnergyPlus are already validated through empirically verifiable data. The discrepancies between simulated and actual building energy performance can be attributed to the inherent uncertainties in the input parameters.
“Robustness” is defined as the ability to tolerate perturbations that might affect the system’s functional body. Robustness of a BEM can be seen as its ability to generate dependable and reproducible energy performance predictions under a wide and diverse range of constructional, technological (electrical-mechanical systems) and operational inputs and under different environmental boundary conditions. Robust BEMs should perform consistently even under varying conditions. EnergyPlus-based whole-building energy models exhibit superior quality over a range of regression based surrogate building models from the point of BEM robustness. One can pose the following fundamental question about EnergyPlus’ BEM robustness:
Why EnergyPlus models are more robust (in terms of modeling demand side building features)?
Computational features of EnergyPlus (E+) program for increased BEM robustness: Computational capacity, multiplicity, connectivity and modularity as explained below:
§ Capacity: E+ is based on a collection of physically-based first-principle algorithms which allow discrete/individual mathematical representations of sub-building systems (such as layer-by-layer definition of an external wall construction or insulated glazing unit). Even though, this approach requires an increased number of input parameters to define a single element, BEM robustness is higher due to the inherent ability to define a wide-range of perturbations of individual building elements which are identifiable/addressable at a much finer resolution compared to alternative lumped models. Enhanced computational capacity of E+ fostering its BEM robustness also manifests itself in its ability to perform temporally flexible and dynamic energy simulations from standard hourly iterations down to the level of minutely iteration cycles. E+ can also characterize building energy performance by generating an extensive and diverse set of simulation outputs at varying levels of frequency. This helps users to identify input-output correlations and variability more comprehensively so as to evaluate model robustness.
§ Multiplicity: E+ allows multiple and alternative definitions of building systems. Such as envelope air infiltration can be defined as air-change-per-hour (ACH) for a thermal zone or as flow rate per unit area of en externally exposed façade element (m3/sec/m2). Glazing assemblies can be defined with very detailed layer-by-layer method (down to solar-optical properties per wavelength of a solar spectrum) or by simply entering overall performance indices for the entire assembly (such as U-factor, SHGC, and VT)[1]. In such a way, E+ can respond to different input definitions (coming from different information sources e.g., manufacturer’s data) in a more robust way. Furthermore, E+ allows its users to switch between different computational algorithms while modeling a building systems behavior. Such as providing alternatives of Conduction Transfer Functions (CTF) versus Conduction Finite Difference (ConFD) methods for surface heat balance calculations. Similar examples can be given for shadow calculations, surface convection calculations, zone air heat balance calculations, etc. It is possible to adjust computational capacity of E+ based on the needs of a project or building data availabilities. These built-in modeling features render E+ as a computationally robust BEM tool for predicting energy performance of new and existing buildings.
§ Connectivity: E+ program is developed with text-based I/O (input-output) data structure. Therefore, E+ is totally open to be coupled with iterative/parametric search routines where individual and detailed building parameters can be defined as elements of a variable/independent parameter sets. Such adaptability allows researchers to run automated design space exploration by varying a finite set of design parameters so as to quantify uncertainties in the model response that leads the way to sensitivity analysis and evaluation of the model robustness for risk analysis. This capability has synergistic interactions with the very first item in this list. Automated and structured parametric search capability achieved via computational connectivity feature of E+ paves the way for development of novel approaches to increase BEM robustness (i.e. parametric analyses for specific building-climate-usage patterns to determine manageable or reducible uncertainties). Furthermore, E+ program can be coupled with external equation-based models in a co-simulation environment (e.g., coupling E+ with mathematical models developed in MATLAB or Modelica via Building Controls Virtual Test-Bed (BCVTB) platform and its successor SOEP project). This computational connectivity feature makes it possible to develop virtual prototypes of advanced-customized-emerging HVAC systems which cannot be handled solely by using E+ program itself. This is surely an advancement for the sake of computational flexibility (responding to variations) resulting in enhanced BEM robustness.
§ Modularity: E+ has a modular and object-oriented structure in its system architecture. For instance, building heating and cooling loads (calculated as a result of demand side building design features) can be decoupled from mechanical systems and can be overwritten by external data inputs (collected from on-site measurements) and energy simulation can continue with E+’s HVAC system simulation routines. This is highly useful when calibrating E+ models which includes larger uncertainties in the inputs of building geometry, envelope constructions, envelope infiltration rates, lighting systems, appliances and occupancy’ profiles. Even individual HVAC system components can be decoupled from the mainstream simulation data flows and some key system parameters can be overwritten by actual data coming from measurements or virtual data from other simulation modules external to E+. This feature also synergistically complements the item listed in the computational connectivity feature. Therefore, computational modularity of E+ helps to improve the BEM robustness with inherent flexibilities coming with the ability of penetrating the data streams within E+ architecture in order to increase fitness to unconventional HVAC system modeling requirements.
[1] U-factor: Overall Coefficient of Thermal Transmission, SHGC: Solar Heat Gain Coefficient, VT: Visible Transmittance.
Computational features that renders EnergyPlus as a robust BEM
EnergyPlus Characteristics in Relation to BEM Robustness in Existing Building Models
It should here be noted that the EnergyPlus program is originally developed as a computational design support tool that is intended to provide quantitative feedback about the whole-building energy performance of a building during the its design phases (before construction and operation). Therefore, different simulation modules of EnergyPlus (HVAC modules) assume ideal system and component design and operation with pre-defined and deterministic targets to reach or satisfy (e.g., heating-cooling thermostat set-points for a thermal zone). While being extremely sophisticated in virtual representations of complete and ideal systems, EnergyPlus has limitations in representations of systems subjected to variable behavior and control sequences with all the uncertainties (under actual operating conditions). EnergyPlus is not a dedicated system control simulation program and may not allow to its users to overwrite simulation data at each and every control point.
However, energy modelers are increasingly using EnergyPlus to develop calibrated energy models of existing buildings that are considered for deep energy retrofit projects in response ever increasing need for characterizing the energy-related and dynamic behavior of existing buildings under actual/real-life operation and ambient conditions. Under such circumstances, and to achieve well calibrated models with EnergyPlus (i.e., achieving predictions between the required thresholds of accuracy), one key point to pay attention is to have a rich set of monitored HVAC system data. The richness in this context is to have multiple time-series data points characterizing system power, as well as water-air mass flow rates as well as temperatures. Such an approach will be crucial in generating calibrated EnergyPlus models with less dependency to its idealistic-deterministic HVAC modeling routines. As a result, provision of diversified and multiple actual system data will consequently diminish the inherent uncertainties in model predictions, hence increase the computational robustness of calibrated models of existing buildings developed with EnergyPlus program.
Note: Jielde Lamp: A modular + mechanical articulating desk lamp designed by Jean-Louis Domecq 1950.
In 1950, a mechanic working in the Lyon region of France set out to create a reading lamp with two goals. Simplicity and Robustness. Jean-Louis Domecq achieved not only that with the "Jielde" lamb, he created a beautiful piece of industrial design, and a unique solution to a wiring problem. Jean-Louis Domecq designed a modular articulating lamp with a very unique innovation. The design of the lamp features no wires at any of the joints. This allows Jielde lamp to rotate without limits on any joint and reduces the risk of electrical shock or fire via borken wiring inside the joint itself.
Omer T. Karaguzel, PhD
Comments