Proceedings of BS2015: 14th Conference of International Building Performance Simulation Association, Hyderabad, India, Dec. 7-9, 2015
PARAMETER IDENTIFICATION FOR LOW-ORDER BUILDING MODELS USING OPTIMIZATION STRATEGIES
Alexander Inderfurth, Christoph Nytsch-Geusen, Carles Ribas Tugores
Berlin University of the Arts Institute for Architecture and Urban Planning Einsteinufer 43-53, 10587 Berlin, Germany firstname.lastname@example.org
The simulation of entire city districts asks for a large number of descriptive parameters for all considered buildings. These buildings are often represented by models using only one thermal zone and few subcomponents in general. This is why parameter sets that describe a district are as dependent on themodel being used as on the actual building.
This article proposes an approach to identify parameter sets for existing buildings based on loworder building models and optimization strategies.
Comparing to a generic, higher order simulation reference and, subsequently, referring to a case study on an existing building the feasibility of this approach is demonstrated.
Besides the simulation of single buildings, simulating the energy performance of entire city districts in order to yield detailed information about their energy-related behaviour gains importance.
Commonly, models used for building performance simulation range from simple one-zone models that might be based on indoor/outdoor temperature differences only, to multi-zone models of great detail, which attempt to describe physical effects with high fidelity. Generally speaking, simple one-zone models require fewer parameters and compute faster but also neglect some of the accuracy inherent to detailed multi-zone models.
Maintaining acceptably short computation times is of great interest when dealing with city districts that might include energy supply and grid models in addition to numerous buildings. Therefore, fast, simplified models are often the tool of choice for the simulation of city districts. Furthermore, city districts display a near infinite
parameter space (Kämpf and Robinson, 2009) to describe each building’s constructions, installations and related user behaviour. Here parameter acquisition and management can be challenging - especially regarding existing building setups where detailed parameter sets of construction materials are often not available. Hence, using simplified models with a reduced parameter count give the additional benefit of a smaller parameterisation effort in district simulations.
In this article the term low-order model is used to describe simplified building models that comprise only a small number of variables dependent on timederivatives
of temperatures, i.e. thermal capacities in the case of building simulations. Low-order models, which focus on simplification, require presumably different parameter sets than more detailed models. That is because, firstly, in most cases more simplified models require fewer parameters altogether. And, secondly, these fewer parameters combined with a model described by an inherently smaller underlying set of equations still need to represent the thermal characteristics of a certain building. Consequently, parameter sets can only provoke optimum results when they are fitted to the model they are actually simulated with. On that account, low-order models described in this article are grey-box models for their reliance on parameter fitting. They contrast with white-box models that are solely based on first principles. Optimization strategies appear to be suitable to fit parameter sets to these low-order models.
This article proposes a strategy that adapts parameter sets to low-order simulation models in order to represent the thermal characteristics of buildings with
known monthly energy consumption and measured weather data. In this context, all modelling is done with the object-oriented modelling language Modelica and simulation with Dymola. The BuildingSystems Modelica library (Nytsch-Geusen et al., 2012) for building performance simulation is used. The GenOpt optimization tool provides various optimization algorithms as well as a profound framework for optimization problems (Wetter, 2001).
The parameter identification strategy is tested against a higher-order reference model, which is also created using Modelica. Comparison with a simulation based
reference with known properties and thermal behaviour allows for an assessment of the proposed strategy's feasibility.
>>> Read the Proceding --->
News Vedi tutte