Prioritizing Building Envelope Retrofits Through Data-Driven Heat Flux Prediction

This repository contains the code, data, and documentation for our workflow to predict surface‑level heat flux in existing buildings and rank envelope components for targeted retrofit. It implements two machine‑learning surrogates (XGBoost and Deep MLP) trained on EnergyPlus simulation outputs of Danish building archetypes, and validates the method on an independent office model.本代码仓库收录了我们用于预测既有建筑表面热通量、并针对围护结构构件开展针对性改造优先级排序的工作流程所涉及的代码、数据集与文档。本工作实现了两类机器学习代理模型（XGBoost与深度多层感知机（Deep MLP）），上述模型基于丹麦建筑原型的EnergyPlus模拟结果完成训练，并在一个独立办公建筑模型上验证了所提方法的有效性。