当前位置: X-MOL 学术Energy Build. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Statistical and machine learning approaches for energy efficient buildings
Energy and Buildings ( IF 6.6 ) Pub Date : 2025-01-13 , DOI: 10.1016/j.enbuild.2025.115309
John A. Paravantis, Sonia Malefaki, Pantelis Nikolakopoulos, Alexandros Romeos, Athanasios Giannadakis, Evangelos Giannakopoulos, Giouli Mihalakakou, Manolis Souliotis

Buildings are responsible for nearly 40% of primary energy consumption. Over the recent decades, numerous methods have been proposed to model, predict, and optimize the heating and cooling energy consumption in buildings, prioritizing efficiency, accuracy, simplicity, and speed. From basic deterministic formulations to advanced machine learning techniques, various methods have been proposed to improve the thermal performance of existing structures and optimize the design of new ones. This manuscript reviews statistical and machine learning approaches in building energy performance simulation, presenting and discussing theoretical considerations and a review of published research studies, covering input, output, distinctive modeling features, and main results. Statistical learning techniques include linear prediction models, generalized linear models, linear mixed-effects models, Bayesian approaches, and time series analysis. Machine learning techniques include deep learning approaches, such as deep feed-forward, recurrent, and convolutional artificial neural networks. Support-vector machines and ensemble machine learning are also discussed, each with a review of relevant research studies, respectively. The application of machine learning approaches in building design and control include both model predictive and reinforcement learning-based control, and building retrofit. The goal is to provide a detailed overview of historical and contemporary developments in data-driven methodologies, encompassing various scientific approaches and algorithms shedding light on the complexities and trends in the dynamic field of energy-efficient building design and operation.

中文翻译:


节能建筑的统计和机器学习方法



建筑物占一次能源消耗的近 40%。近几十年来,人们提出了许多方法来建模、预测和优化建筑物的供暖和制冷能耗,优先考虑效率、准确性、简单性和速度。从基本的确定性公式到先进的机器学习技术,已经提出了各种方法来提高现有结构的热性能并优化新结构的设计。本手稿回顾了建筑能源性能模拟中的统计和机器学习方法,提出和讨论了理论考虑因素,并回顾了已发表的研究,涵盖输入、输出、独特的建模特征和主要结果。统计学习技术包括线性预测模型、广义线性模型、线性混合效应模型、贝叶斯方法和时间序列分析。机器学习技术包括深度学习方法,例如深度前馈、递归和卷积人工神经网络。还讨论了支持向量机和集成机器学习,分别回顾了相关的研究。机器学习方法在建筑设计和控制中的应用包括基于模型预测和强化学习的控制,以及建筑改造。目标是详细概述数据驱动方法的历史和当代发展,包括各种科学方法和算法,阐明节能建筑设计和运营动态领域的复杂性和趋势。
更新日期:2025-01-13
down
wechat
bug