|
| 办公街区多尺度设计要素与建筑碳排放的非线性关系研究——基于参数化与SHAP方法 |
| A Research of the Non-linear Relationship between Multi-scale Design Elements and Building Carbon Emissions in Office Blocks——Based on Parameterisation with the SHAP Method |
| 投稿时间：2024-06-01 修订日期：2024-09-04 |
| DOI：10.12285/jzs.20240601001 |
| 中文关键词: 办公街区、多尺度设计要素、建筑碳排放、非线性关系、参数化、机器学习、SHAP |
| 英文关键词: Office blocks, Multi-scale design elements, Building carbon emission, non-liner relationship, Parametric design, machine learning, SHAP |
| 基金项目: |
|
| 摘要点击次数: 110 |
| 全文下载次数: 2 |
| 中文摘要: |
| 碳中和城市街区推广建设与数智时代背景下，采用机器学习算法实现城市街区的建筑碳排放快速预测以及背后影响机制的解读可以辅助建筑师在方案设计阶段进行智能决策。针对机器学习预测模型解释性不足的问题，本文旨在构建一种基于参数化与SHAP方法的办公街区多尺度设计要素与建筑碳排放的非线性关系分析范式。基于武汉市办公街区案例调研建立参数化模型，采用6种集成学习算法构建了办公街区建筑碳排放预测模型，其中Gradient Boosting （梯度提升）算法的性能最佳。以此模型为例，进行办公街区多尺度设计要素与建筑碳排放的非线性关系解读，识别影响建筑碳排放的关键设计参数，解读背后的影响机制，辅助建筑师进行设计方案的智能决策。 |
| 英文摘要: |
| In the context of promoting the construction of carbon-neutral urban blocks and the digital intelligence era, the use of machine learning algorithms to achieve rapid prediction of building carbon emissions in urban blocks and the interpretation of the impact mechanisms behind them could assist architects in making intelligent decisions at the pre-design stage of a scheme. Aiming at the problem of insufficient explanatory nature of machine learning prediction models, this paper aims to develop a paradigm for analysing the non-linear relationship between multi-scale design elements and building carbon emissions of office blocks based on parameterisation and SHAP methods. A parametric model was established based on the Wuhan office block case study, and six ensemble learning algorithms were used to construct a building carbon emission prediction model for office block, among which the gradient boosting algorithm had the best performance. This model is used as an example to interpret the non-linear relationship between multi-scale design elements and building carbon emissions in office blocks, identify key design parameters that affect building carbon emissions, and interpret the influencing mechanisms behind them, so as to assist architects in making intelligent decisions on design solutions. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |
