Dynamic Simulation

Dynamic simulation of buildings, also called Building Performance Simulation (BPS), is generally accepted as a powerful tool for analysing the thermal and energy performance in buildings (Waltz, 2000; Clarke, 2001; Davies, 2004).

The use of computer-aided design tools was firstly introduced into architectural and engineering practices in early 1960s. The energy crisis of 1970s resulted in the development of several (initially simplified) computer-based building energy performance prediction tools (Raslan,

(Clarke & Hensen, 2015). There are currently two main approaches to building energy modelling:

• Simplified modelling methods

• Complex dynamic simulation methods

The former category includes simplified approaches to building modelling that are either: i. Steady-state calculation methods, using variables averaged over a longer period of time

(monthly, seasonally or annually). These models involve certain assumptions to the underlying model of the building. Several energy flow-paths, usually dynamic in nature, are approximated or neglected completely. They are commonly used for fast and low- cost estimation of building performance, for benchmarking and comparing a building to a “stock average” building of the same type.

ii. Simplified dynamic methods are often used to demonstrate compliance with building regulations (as per CEN standards) (Kokogiannakis et al., 2008). The simplified dynamic models take into account the effect of transient parameters (such as weather) to achieve more accurate predictions of building performance (Kim & Kim, 2007). The simplified modelling methods, also referred to as “calculation tools” (Raslan, 2010), do not aim to investigate all complex interactions between the building and the surrounding environment, in contrast to the other category of dynamic simulation tools. The tools that fall into the category of dynamic simulation methods account for all possible energy flow-paths and their interactions within a building (Clarke, 2001). They involve complex and iterative predictive analytical procedures and they typically use hourly or sub-hourly time steps (Raslan, 2010). The dynamic tools take fully into account the transient performance of the building and they are considered more realistic and more accurate in predicting the overall energy

performance of design proposals. A thorough literature review on the benefits and challenges associated with the use of dynamic BPS can be found in Section 2.3 of this thesis. In the context of this project, only dynamic simulation tools have been used to quantify the energy consumption and the thermal performance of a simple ICF building model and to address the requirements of Objective No1.

3.3.2.1

Comparative Testing of BPS Tools

An inter-model comparative testing was employed as the key research method in the first stage of the research (i.e. computational analysis). This step was focused on reviewing and contrasting the main features and capabilities of a list of nine widely-used BPS tools and evaluating their ability to predict the thermal performance of ICF using whole BPS. The building model selected for this step of the analysis was a single-zone test building based on the one specified in the BESTEST methodology (Judkoff & Neymark, 1995).

As discussed in Section 2.3.4, comparative testing is a common validation method used to compare a simulation program to itself or to other programs (Judkoff & Neymark, 1995; Ryan & Sanquist, 2012). Its main limitations involve the lack of an absolute truth and the assumption that the other models are accurate and validated. Hence, prior to proceeding to any comparison, it was essential to verify that the models used for the analysis were “validated” and thus capable of delivering reliable results. International Energy Agency (IEA) Building Simulation Test (BESTEST) and diagnostic method (Judkoff & Neymark, 1995), also adopted in ASHRAE Standard 140 (ANSI/ASHRAE, 2014), was used for model validation. The BESTEST method consists of a number of building energy simulation test (BESTEST) suites and it is used for evaluating the modelling capabilities of whole building performance simulation tools and for diagnosing errors in their source code. The output data from a number of widely-used BPS tools (state-of-the-art) are provided as a basis for comparison and are used to define an “acceptable”

case 600 (low thermal mass) and case 900 (high thermal mass) were used to validate all simulation models (from all nine BPS tools) and to ensure that there are no input errors that could lead to significant inaccuracies in the results. Then the construction details were changed in line with the specific study. To ensure consistency, all other input parameters remained identical to the BESTEST methodology. Moreover, all simulations were performed by the author to minimise the influence of user variability on the results (Guyon, 1997; Berkeley et al., 2014).

3.3.2.2

Default Models

The simulation results (for the same single-zone building) provided by the nine BPS tools were compared to each other, relying initially on the default settings and solution algorithms employed by the various tools. The divergence of the nine tools was investigated by looking at the annual heating and cooling energy consumption and the annual peak heating and cooling loads. This variability was analysed by means of percentage difference between minimum and maximum values (to show the range of variation), and by looking at the percentage difference of each individual tool from the median of all tools. This step provided some insight into the level of modelling uncertainty associated with the simulation of ICF in buildings. Further details can be found in the papers in Appendix A and B.

3.3.2.3 Model “Equivalencing”

To identify key parameters and modelling factors that contribute to modelling uncertainty when simulating an ICF building, two of the nine BPS tools (which showed relatively consistent results in the first instance of the analysis) were selected for further investigation (Tools E and I). Monthly and hourly predictions on heating and cooling energy consumption, system loads and surface temperatures were analysed and compared with the use of Normalised Root Mean Square Error (NRMSE). The NRMSE is a metric used to quantify the typical size error between sets of data relative to their mean value (Granderson & Price, 2013). For example, a 10%

NRMSE means 10% difference from the mean value. The NRMSE when normalised to the mean of the observed data is also called CV-RMSE. In the model “equivalencing” process the use of the NRMSE was selected to avoid any confusion with the CV-RMSE as defined in the ASHRAE 14 Guidelines for model calibration (ASHRAE, 2014). In the ASHRAE 14 Guidelines the denominator is the mean of the measured energy data. In the analysis reported here the denominator was the mean value of the simulation predictions provided by the two BPS tools. The equation of the NRMSE is given below.

𝑁𝑅𝑀𝑆𝐸 (%) = √ ∑𝑛_{𝑖=1(𝑥𝑖,𝑖− 𝑥𝑖,𝑒)2} 𝑛 𝑥̅ ∗ 100 (Eq.5) 𝑥̅ = _𝑖 𝑥𝑖,𝑒+ 𝑥𝑖,𝑖 2 (Eq.6) 𝑥̅ = ∑𝑛𝑖=1𝑥̅𝑖 𝑛 (Eq.7) Where,

𝑥𝑖,𝑒 and 𝑥𝑖,𝑖 are the predictions provided by tools E and I respectively at each time step 𝑥̅_𝑖 is the mean value of 𝑥_𝑖,𝑒 and 𝑥_𝑖,𝑖 for each time step

𝑥̅ is the mean value of the predictions provided by both tools E and I

n is the size of the sample

The aim of this step was to reflect on the impact that the various solution algorithms and calculation methods had on the variability of results. To achieve this, a step-wise method of

differences between the two tools by changing to identical algorithms and simulation settings, where possible. Details of the model “equivalencing” method can be found in the paper in Appendix B.