A number of studies have been undertaken in order to assess the sensitivity and uncertainty of the input parameters within building energy assessment models.
Due to the need to standardise procedures in order to ensure consistency across the evaluation of different buildings by individual assessors, several simplifications and assumptions are used, in addition to fixed calculation methods. This can lead to uncertainty in the model itself, which can impact upon the ability of the model to provide truly accurate outputs (de Wita et al., 2002). Figure 2-11 summarises the main techniques used in sensitivity analysis.
Figure 2-11 - Sensitivity Analysis - Main Techniques
Produced from Source Data: (Frey et al., 2003; Kavgic et al., 2010; Saltelli et al., 2000)
Following a detailed study of input parameters in building energy models, Lomas (1992) recommended that DSA was most suitable for evaluating uncertainty in individual input parameters (local impact), with MCA providing a better tool for assessing total/overall sensitivity of the model to cumulative
changes in parameters (global impact). However, whilst the practise of varying one parameter whilst fixing all others (one factor at a time DSA) is commonplace in the evaluation of building energy models, there is concern that this may not provide a reliable assessment of the accuracy of the model (Saltelli et al., 2010). This is largely due to the assumption within DSA that all inputs are independent and have no impact upon one another, which is clearly not the case in the context of building performance (Lomas et al., 1992).
MacDonald (2002) and Booth (2012) have assessed the aspects of building energy models that may potentially lead to inaccurate analysis of building energy demands, and suggest that the main areas of concern are:
Ability of the model to represent reality;
Accuracy and appropriateness of derived/assumed data in the absence of true measured data for input parameters;
Assumptions regarding climate, occupancy and behaviour, system installations and use;
Appropriateness of default values, fixed parameters and base calculations;
Accuracy of data input; and
Effect of late design changes on final predictions.
Detailed assessment of the UK BREDEM model has also been undertaken in response to concerns regarding the robustness of the model, including many of the issues identified previously in more general studies. Palmer (2012) used a one at a time DSA approach in order to ascertain the parameters in BREDEM 9 that can have the most significant impact on final outcomes.
Table 2-1 (Palmer et al., 2012, p. 137) shows, in descending order, the most influencing factors, ranked using a normalised sensitivity coefficient. These values reflect the absolute effect of a change in each parameter on the final calculated energy consumption. For example, a 1°C rise in internal demand temperature will result in a 1.54% rise in energy consumption. This study suggests that many of the highest ranking sensitive parameters form part of the heat losses and ventilation section within the model (as highlighted by shading in Table 2-1).
Table 2-1 - Significant Parameters Identified in BREDEM 9 Following Differential Sensitivity Analysis
Source: (Palmer et al., 2012, p. 137)
Input Parameter Initial Set
Internal Demand Temperature(°C) 19.0 1.54 Main Heating System Efficiency (%) 80.5 -0.66
External Temperature (°C) 7.5 -0.59
Total Floor Area (m2) 96.4 0.53
Storey Height (m) 2.5 0.46
Daily Heating Hours (hrs) 11.0 0.27
DHW System Efficiency (%) 76.6 -0.19
Wall U-value (W/m2K) 1.2 0.18
Effective Air Change Rate (ach) 1.0 0.18
Wind Factor Parameter 4.0 -0.17
Wind Speed (m/s) 4.8 0.17
Infiltration Rate (ach) 0.8 0.17
Appliances Energy Coefficient (TFAxN) 0.47 0.17
Shelter Factor 0.9 0.16
Main Heating Responsiveness 0.9 -0.15
Total 998 908
Similarly Quigley (2010), undertook a detailed sensitivity analysis of BREDEM 8 as applied to several case study building scenarios, and concluded that fabric u-values, air permeability data and heating technologies had the greatest influence on final outputs. Work undertaken by Firth et al. (2010) also found that the characteristics of heating systems and building heat losses had the most impact on energy demands and carbon emissions calculated by a bespoke base model derived from a BREDEM 8 foundation.
Kavgic (2010) observes that the main limitations of building energy models are their lack of transparency due to hidden algorithms, inability to alter certain data inputs and outputs, and uncertainty surrounding assumptions used. The evidence suggests that, whilst the limitations of BREDEM and subsequently SAP, are acknowledged, actual quantification of the impact of individual parameters on final outputs is limited.
Within SAP, there are a large number of input parameters, assumptions and underlying calculation formulae that ultimately influence the output data. It can be seen from the studies undertaken to date that fabric heat losses and ventilation rates are identified as some of the most significant areas of the model in terms of their ability to affect final energy demand values.
There are a number of key items within the SAP 2009 methodology that contribute to the calculation of a HLC output value, as shown in Figure 2-12.
This provides a measure of the whole house heat losses in terms of W/K, that is, the required energy (W) required to heat the building per degree of difference between the internal and external temperatures (K).
The calculation is derived from BS EN ISO 12831 (Building Standards Institute (BSI), 2003), where Total Design Heat Loss is equal to the sum of the design transmission heat loss for heated space (W) and design ventilation heat loss for heated space (W). Element u-value multiplied by element surface area data provides a value of fabric heat loss for each aspect of the building (floors, walls,
background infiltration and additional ventilation losses are summed, together with fabric heat loss, to calculate the HLC (W/K) under steady state conditions.
This provides a measure of the required energy (W) required to heat the building per degree of difference between the internal and external temperatures (K).
Figure 2-12 - Heat Loss Coefficient Calculation - Key SAP Input and Output Data (Produced by Author)
The HLC is used within SAP methodology to calculate space heating requirements, which account for two thirds of total energy demand in an average UK home (as illustrated previously in Figure 1-6). Therefore, this parameter requires careful calculation in order to accurately predict the energy consumption of a dwelling. As a benchmark for fabric performance of buildings, the HLC has been identified as an appropriate means to evaluate thermal efficiency of housing within this research.