- The paper applies a combined first and second-order elementary effects method (Morris method) for sensitivity analysis to building energy simulations, evaluating parameter influence and interactions.
- The analysis identifies critical input parameters such as insulation thickness and set-point temperature that significantly impact heating demand, distinguishing linear and non-linear effects.
- The findings assist modelers in prioritizing data collection and refining computational models while advancing the understanding of parameter interactions in complex systems.
Sensitivity Analysis in Building Energy Simulations: A Combined Approach
The paper "Application of sensitivity analysis in building energy simulations: combining first- and second-order elementary effects methods" explores the utilization of sensitivity analysis techniques to enhance the accuracy and efficiency of building energy models. The research leverages sensitivity analysis, specifically the Morris method, and its extension to second-order interactions, to evaluate and improve the modeling of energy demand for a multi-zone apartment building using the ESP-r simulation software.
The sensitivity analysis is pivotal in determining the influence of various input parameters on the outputs of complex models, such as those predicting building energy consumption. This paper highlights the usefulness of combining first- and second-order sensitivity analyses to garner insights into both linear and non-linear influences of parameters, as well as their interactions that might otherwise go unnoticed using conventional methods.
The authors begin with a robust review of existing sensitivity analysis methodologies. They categorize these into techniques evaluating individual parameter influence, those using random sampling methods, and methods involving segmented input distributions. The distinction between local and global methods is underscored, with global methods posited as more comprehensive for encompassing a wider domain of input variables.
Central to the analysis is the use of the Morris method, a distinct approach designed to prioritize input factors based on their elementary effects or first-order derivatives. The paper extends Morris's method to assess second-order interactions, thereby increasing the depth of analysis by accounting for parameter-pair interactions. This is particularly imperative in complex systems like building thermal models, where interactions can significantly alter outputs.
In applying these methods, the authors use a seven-storey apartment building as a case paper, analyzing multiple parameters affecting the thermal simulation. The results distinguish critical input factors such as building dimensions, set-point temperature, ventilation rate, and insulation thickness, which are shown to have significant impacts on heating demands and comfort levels. The analysis convincingly delineates between linear parameter effects and those exhibiting non-linearity or interaction effects, guided by specific numerical thresholds for determining non-linearity.
The first-order analysis demonstrates the capacity of the Morris method to efficiently identify primary influential parameters, while the second-order analysis reveals complex parameter interactions and non-linear behavior. The latter is of great interest, as interactions between parameters (e.g., building size-related parameters and set-point temperature) present unique challenges and opportunities for model simplification and refinement.
By employing different output transformations, such as converting annual heating needs into logarithmic scales, the paper successfully reduces parameter correlation and illuminates multiplicative interactions. This transformation appears beneficial in emphasizing certain parameter effects that might be obscured otherwise.
The implications of this research are significant, underscoring the potential of sensitivity analysis for improving building energy models. Practically, it assists modelers and engineers in prioritizing data collection and refining computational models. Theoretically, it advances the understanding of parameter interactions in complex systems, which may inspire further research into computationally efficient methods for even higher-order sensitivity analyses.
Future developments in sensitivity analysis would benefit from incorporating even more advanced statistical techniques and exploring machine learning integration to further reduce computational costs and enhance prediction accuracy. The comprehensive methodology employed serves as a blueprint for sensitivity analysis applications beyond building energy simulations, potentially impacting a range of disciplines where complex stochastic models are prevalent.
Overall, the paper provides a detailed and methodologically sophisticated exploration of sensitivity analysis in building energy modeling, showcasing how combining first- and second-order methods can significantly enhance the understanding and predictive capabilities of these models.