A GUIDE TO BAYESIAN CALIBRATION OF BUILDING ENERGY MODELS
A GUIDE TO BAYESIAN CALIBRATION OF BUILDING ENERGY MODELS
复制标题
建筑能源模型贝叶斯校准指南
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
R. Muehleisen
中科院分区：
文献类型：
--
作者：
Matthew E. Riddle;R. Muehleisen
贝叶斯校准提供了大量的好处，超过标准的方法来校准建筑物能源模型。它提供了一个自动化的过程，优化参数值，不严重依赖于专家知识。它还考虑了多个不确定性来源，以生成对可能结果范围的估计，而不是单个模型预测。然而，使用贝叶斯校准可能令人望而生畏。本文旨在通过概述设置和使用贝叶斯校准的关键步骤以及解决每个步骤中的设计问题，使其更易于使用。关键步骤包括量化不确定性，筛选参数，进行模型运行以生成元模型，并根据模型运行的结果和测量的现场数据执行贝叶斯校准。在每一步中，我们都将探索必须做出的关键决策以及与使用不同方法相关的权衡。我们还讨论和比较了可用于帮助贝叶斯校准的软件包。
Bayesian calibration provides substantial benefits over standard approaches to calibrating building energy models. It provides an automated process for optimizing parameter values that does not rely heavily on expert knowledge. It also accounts for multiple sources of uncertainty to generate an estimate for the range of possible outcomes rather than a single model prediction. However, using Bayesian calibration can be daunting. This paper aims to make it more accessible by outlining the key steps in setting up and using Bayesian calibration and by addressing design issues at each step. Key steps include quantifying uncertainty, screening parameters, conducting model runs to generate a metamodel, and performing Bayesian calibration based on the results of the model runs and measured field data. At each step, we explore the key decisions that must be made and the tradeoffs associated with using different approaches. We also discuss and compare software packages that can be used to aid in Bayesian calibration.