Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


Uncertainty analysis of building energy analysis based on replicated Latin Hypercube sampling

Wei Tian 1, Miaorou Jin 1, Pieter de Wilde 2, Xing Fu 3, Guangchen Li1
1 Tianjin University of Science and Technology, China
2 Plymouth University, UK
3 Tianjin Architectural Design Institure, China

Abstract: Uncertainty analysis has become increasingly important to provide reliable energy efficiency assessment of buildings whilst considering inherent uncertain factors. The one-batch sample method is usually used with the Latin Hypercube sampling (LHS) to evaluate building performance, however it is difficult to assess the stability of uncertainty results using this approach. Hence, the uncertainty of energy performance in buildings may be exaggerated or understated. This paper explores the application of replicated LHS in order to properly evaluate the stability of uncertainty results in building energy assessment. The results indicate that the replicated LHS can quantify the stability of energy analysis by obtaining the confidence intervals of energy use. The similarity of replicated results can be analysed using formal statistical distance measures, including Kolmogorov-Smirnov distance and Dowd test statistics. Moreover, the combined results from several repetitions can improve the convergence of uncertainty results in building energy analysis.
Keywords: Building energy analysis, uncertainty analysis, Latin hypercube sampling, confidence interval
Pages: 1303 - 1310