.png){kind=icon}
返回
Causality-driven, interpretable surrogate modeling for DOAS preheating control
.png){kind=icon}
PRE
AI
DOI：10.1016/j.enbuild.2025.116625.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Dedicated outdoor air systems (DOAS) that supply 100% outdoor air require reliable, model-based preheating control to prevent freezing and reduce energy consumption. Although physics-based models provide equationdriven rigor and transparency, their complexity often hinders intuitive causal insights, leaving operators without actionable guidance for real-time control. To bridge this gap, this study introduces a causality-driven surrogate model, or a linear double machine learning (DML) simulator, for the preheating coil (PHC) in a DOAS. In the DML formulation, the effect of the control variable (valve opening ratio) on the outcome (outlet air temperature) is estimated via successive residual-on-residual regressions that account for covariates - such as outdoor air temperature and hot water inlet temperature. As the data source for the surrogate model, a highfidelity physics-based simulator was used after being calibrated with field measurements from a semiconductor-facility. Synthetic data generated from this model were then used to train the DML surrogate, enabling it to represent the underlying physical causal relationships for predicting PHC outlet air temperature. The resulting surrogate model demonstrates predictive accuracy on par with the full simulator across a range of conditions, indicating robust generalization (mean absolute error: 1.1 degrees C for measured data; 1.4 degrees C for synthetic data). Its low-dimensional numerical form produces interpretable coefficients that capture the context-dependent causal impact of the valve opening ratio, providing actionable insights into energy trade-offs between boiler heating and pump operation. This combination of predictive fidelity and causal interpretability enables efficient real-time optimization of PHC control.
Keyword:
Dedicated outdoor air system (DOAS)
Preheating coil
Heating control
Surrogate modeling
Interpretable model
Causal inference
Double machine learning (DML)
Real-world building data
期刊
IF:
7.1
论文数:
1.6W
被引数:
6.8W
机构
引用论文
Ten questions concerning model predictive control for energy efficient buildings关于节能建筑模型预测控制的十个问题
PREAI
分享收藏Optimization-informed rule extraction for HVAC system: A case study of dedicated outdoor air system control in a mixed-humid climate zoneHVAC系统的优化通知规则提取: 混合湿润气候区专用室外空气系统控制的案例研究.png)
ENERGY AND BUILDINGS
OAAI
ENERGY AND BUILDINGSIF7.1
OAAI分享收藏Local vs. federated cooling control for an office space with heat pump and photovoltaic systemsENERGY AND BUILDINGS
ENERGY AND BUILDINGSIF7.1
PREAI
分享收藏Choi, Y., et al., Large-scale energy cost optimization and performance analysis for dedicated outdoor air system: simulation results from ASHRAE RP-1865. Science and Technology for the Built Environment, 2024. 30(10): p. 1217-1235.Choi, Y., et al., 大规模能源成本优化和专用室外空气系统的性能分析：ASHRAE RP-1865的模拟结果。建筑环境科学与技术，2024. 30(10): 第1217-1235页。
IF0
PREAI
分享收藏Energy-efficient operation of a complete Chiller-air handing unit system via model predictive control通过模型预测控制实现完整的冷水机组-空气处理单元系统的节能运行
PREAI
分享收藏Multi -zone field study of rule extraction control to simplify implementation of predictive control to reduce building energy useENERGY AND BUILDINGS
ENERGY AND BUILDINGSIF7.1
PREAI
分享收藏Extracting interpretable building control rules from multi-objective model predictive control data setsENERGYOAAI
ENERGYIF9.4
OAAI分享收藏
查看更多内容