Volume 66

Evaluation of an Integrated Fault Detection and Compensation for Building-Level Radiator Systems in District Heating Nima Monghasemi, Amir Vadiee

Abstract

Valve and control faults in building-level radiator systems degrade the efficiency of district heating networks and cause occupant discomfort. Although methods for fault detection and compensation exist independently, integrated systems that diagnose and react to specific faults are a recognized research gap. This study presents an integrated fault detection and compensation framework implemented and evaluated within the EnergyPlus simulation environment. The system uses a machine-learning-informed proxy to trigger fault-specific supervisory control actions. A comparative analysis of control-based (stuck-open) versus physically constrained (stuck-closed) faults through an ablation study revealed fundamentally different operational behaviors. For the stuck-open fault, the fault detection and compensation system was highly effective, reducing the discomfort degree-hours by 68% and the energy consumption by 23%. For the stuck-closed fault, the system’s inability to restore comfort confirmed its role as a high-fidelity diagnostic tool, proving that such physical faults require maintenance rather than supervisory control. These findings establish a framework for fault-specific responses in district heating-connected buildings.

Keywords District heating, fault detection and diagnostics, hydronic heating systems, machine learning