These reports may be ordered, free of charge,
from the Simulation Research Group at
Lawrence Berkeley National Laboratory. Please contact
Kathy Ellington
MS: 90-3147
Lawrence Berkeley National Laboratory
Berkeley, CA 94720
USA
Fax: 510-486-4089
Email:
info@gundog.lbl.gov
LBL-28197 -- INTEGRATION OF SIMULATION INTO BUILDING DESIGN: THE NEED FOR A JOINT APPROACH,
Abstract:
We describe the need for a joint effort between design researchers and simulation tool
developers in formulating procedures and standards for integrating simulation into the
building design process. We review and discuss current efforts in the U.S. and Europe
in the development of next-generation simulation tools and design integration techniques.
In particular, we describe initiatives in object-oriented simulation environments
(include the U.S.'s SPARK, the U.K.'s Energy Kernel System, the Swedish IDA system,
and the French ZOOM program) and consider the relationship of these environments to
recent R&D incentives in design integration methodology provided by the Commission of
the European Community (the COMBINE program). Topics discussed include the role of
simulation in building design, deficiencies of current energy performance evaluation
tools, characteristics of intelligent building design systems, transfer of data and
knowledge between simulation and design, and the STEP standard for the exchange of
product model data.
LBL-28272 -- OBJECT-ORIENTED PROGRAMMING, EQUATION-BASED SUBMODELS, AND SYSTEM REDUCTION IN SPARK,
Abstract:
Collaborative efforts among building simulation researchers in Europe and the U.S. have
resulted in wide acceptance of certain features as necessary attributes of future
simulation environments. As identified in the Energy Kernel System (EKS), the principal
features are those of the object-oriented programming (OOP) paradigm, in which a hierarchy
of submodels is readily defined and interconnected to form system models of widely varying
purpose, solution methodology, and implementation design. The Simulation Problem Analysis
and Research Kernel (SPARK) is an early, prototypical implementation intended to exhibit
at least some of the EKS features, including an OOP-like environment. Additionally, SPARK
implements a solution process that is based on graph algorithms and achieves solution
efficiency through a reduction in size of the iterative problem. This paper enlarges on
earlier descriptions of SPARK, attempting to separate and clarify the modeling issues
and solution issues. To this end, the nature of the SPARK environment is compared to
established OOP environments. The SPARK equation-based objects are contrasted to the
assignment-based (input/output oriented) models employed in most current simulation
environments. The SPARK cut-set reduction methodology is compared to the more widely-known
sparse matrix technique. Finally, the utility of the SPARK reduction algorithm is
demonstrated on a number of example problems.
LBL-28274 -- NEUTRAL FORMAT AND AUTOMATIC TRANSLATION FOR BUILDING SIMULATION SUBMODELS,
Abstract:
Much current research is directed towards development of software environments that
allow easy construction of building simulation models of widely varying structure and
purpose. The example, TRNSYS, has been in use for a number of years. Recently, several
new such environments have been proposed. In spite of a considerable variation in
model description formats among environments, the underlying mathematical models of
physical processes are very similar. While one of the principal aims has been to
allow easy sharing of models among users of the same environment, it has not been
clear how portability was to be provided among different environments. Another
objective has been ease of component model definition, in order to encourage
modifications and additions to model libraries. This paper addresses both of these
issues, by proposing a neutral and natural format for component model expression.
The proposed format encourages equation based model definition because such models
can be converted to efficient algorithmic form if needed, whereas the converse is
not always true. Nonetheless, algorithmic component descriptions are also supported
in order to allow reuse of existing models. Other key features of the proposed format
are typing and declaration of linkage elements between models, which allow development
of compatible component families, and enhance submodel exchange and reuse. The
proposal considers underlying system modeling issues, including hierarchical submodel
decomposition and methods for formal model expression that allow automatic translation
to various simulation environments. Also discussed are the software tools needed for
library maintenance and model translation.
LBL-28275 -- SPECIFYING DYNAMIC MODELS IN THE SIMULATION PROBLEM ANALYSIS KERNEL
Abstract:
The Simulation Problem Analysis and Research Kernel (SPARK) is a prototype
implementation of the Energy Kernel System (EKS). It is intended to demonstrate
some of the features that have been argued as necessary for future simulation systems.
In particular, with earlier versions of SPARK, the model designer can specify static
behavior as modeled by systems of algebraic equations. More recent work has led to
extensions that allow the designer to specify dynamic behavior as modeled by systems
of equations together with certain predefined integration methods. One drawback of
this approach is that the integration methods must be of a specific kind, involving
at most a fixed number of values from previous time steps. In this paper, we propose
to extend SPARK in a more flexible way to give the designer the ability to specify
any of the standard difference methods for initial value problems. These include
Runge-kutta methods as well as predictor-corrector methods. The development of our
extensions begins with a close look at how the static version of SPARK specifies
models. On this foundation, we then look at two very simple integration methods:
Euler's implicit and explicit methods for clues as to how to can extend the system
description language to admit dynamics. With the extensions to SPARK called for by
looking at Euler's methods, we show that Runge-Kutta and predictor-corrector methods
can also be specified. From this, we argue that the extensions proposed herein are
suitable for specifying arbitrary finite difference integration methods. We conclude
with a brief discussion of further extensions that we believe would make the SPARK
system more useful; for dynamic simulations.
LBL-21144 THE EXPONENTIAL SCHEME FOR COMPUTATION OF NATURAL CONVECTION FLOW IN ENCLOSURES,
Abstract:
In this paper the exponential scheme for multi-dimensional unsteady problems is discussed,
a basis for approximation of solution is given, and a boundary exponential scheme is
derived. It is applied to the computation of natural convection in a square cavity with
moderate size uniform grid for future calculation of air flow in enclosures and heat flux
at walls in building energy analysis. The numerical method for the Boussinesq equations is
based on the Marker and Cell (MAC) method and is put into conservation form via the
Spalding-Patankar flux. The latter method is also discussed. Preliminary numerical
tests show that the method is promising.
LBL-23617 -- DAYLIGHTING DESIGN FOR THE PACIFIC MUSEUM OF FLIGHT: ENERGY IMPACTS
Abstract:
The daylighting performance of the Pacific Museum of Flight in Seattle, WA, has been
analyzed using the DOE-2 building energy simulation program. The main exhibit areas of
this museum are enclosed on three sides by glass walls and the 48,000-sqft roof is
completely glazed. Because of the large glass areas, a detailed thermal simulation of
the building was carried out during its design phase in order to select glazing parameters
that would avoid excessive summer solar heat gain, reduce winter heat loss and, at the same
time, provide enough natural light to significantly reduce electric lighting loads.
Glazing choices considered included conventional glass, heat mirror, and glass with a
low-emissivity coating. On/off, stepped, and continuous dimming lighting control systems
were analyzed. Daylighting was found to be very effective in reducing annual electric
lighting load, peak electrical demand, and the overall annual energy consumption.