Sustainable subway indoor air quality monitoring and fault-tolerant ventilation control using a sparse autoencoder-driven sensor self-validation

• Published in: Sustainable cities and society Vol. 52; p. 101847
• Main Authors: Loy-Benitez, Jorge, Li, Qian, Nam, KiJeon, Yoo, ChangKyoo
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-01-2020
• Subjects: Fault-tolerant ventilation control; IAQ monitoring; Sensor self-validation; Sparse autoencoder; Subway transportation systems; Sensor self-validation; Sparse autoencoder; IAQ monitoring; Fault-tolerant ventilation control; Subway transportation systems
• ISSN: 2210-6707; 2210-6715
• DOI: 10.1016/j.scs.2019.101847

•An AE sensor self-validation method for IAQ sensors in subway stations was proposed.•The influence of faulty and reconstructed IAQ over the health risk was assessed.•The AE method showed superiority on sensor validation over typical methods.•A fault tolerant ventilation control was obtained from the AE reconstruction.•Sustainable IAQ ventilation control system was modeled using AE reconstruction.
Sensors providing measurements for monitoring and control of indoor air quality (IAQ) are key components of the ventilation systems in subway stations. However, faulty sensors due to harsh ambient conditions may deliver incorrect information triggering misinterpretation; causing energy waste or IAQ deterioration. This paper presents a holistic online framework for sensor self-validation in a subway station based on a sparse autoencoder (AE) architecture. The sensor self-validation procedure consists of sensor fault detection, faulty sensor identification, and faulty sensor reconstruction. First, the AE-based detection rate between 44% and 100%. Then, the faulty sensor identification was conducted through an AE-sensor validity index (SVIAE). The faulty sensor reconstruction was conducted by the AE structure and evaluated with several performance metrics. Finally, the sustainability and fault-tolerance aspects of this framework were verified through mathematical modeling of the ventilation system; showing the effects of the faulty and reconstructed sensors on energy consumption and public health.