Abstracts:
With the growing urbanization process in several cities around the world, air pollution mitigation has become one of the main environmental challenges of the present time. Recently, low-cost air pollution sensors have become a current trend in the air quality control area since they are an affordable alternative for deploying air quality monitoring systems with high spatial resolution. However, the main drawback of these sensors is that they tend to provide measurements with lower accuracy and reliability compared to traditional air quality monitoring stations. Therefore, periodical calibration of these sensors is essential to maintain the quality of their measurements. This work presents a novel air quality sensor calibration method based on a Bayesian neural network model. The proposed method is assessed using a real public available dataset. The test experiment results show that the method has a good accuracy performance, with a lower mean absolute error compared to other machine learning-based calibration methods applied to the same dataset. In addition, the method presents the advantage of directly providing estimations of the uncertainty of the calibrated measurements, which is an important metric used to assess the quality and reliability of data provided by air pollution sensors and that most other calibration methods usually cannot provide
Reference:
TAIRA, Gustavo R.; LEAL, Adriano Galindo; SANTOS, Alessandro Santiago; PARK, Song W. Bayesian neural network-based calibration for urbana ir quality sensor. In: EUROPEAN SYMPOSIUM ON COMPUTER AIDED PROCESS ENGINEERING, ESCAPE 32, 2022, Toulouse , France. Proceedings… 6 p.
Document with restricted access. Access to the work in PDF, only for users registered in the library. Request the password at the DAIT Library Service:
https://escriba.ipt.br/pdf_restrito/178092.pdf
With the growing urbanization process in several cities around the world, air pollution mitigation has become one of the main environmental challenges of the present time. Recently, low-cost air pollution sensors have become a current trend in the air quality control area since they are an affordable alternative for deploying air quality monitoring systems with high spatial resolution. However, the main drawback of these sensors is that they tend to provide measurements with lower accuracy and reliability compared to traditional air quality monitoring stations. Therefore, periodical calibration of these sensors is essential to maintain the quality of their measurements. This work presents a novel air quality sensor calibration method based on a Bayesian neural network model. The proposed method is assessed using a real public available dataset. The test experiment results show that the method has a good accuracy performance, with a lower mean absolute error compared to other machine learning-based calibration methods applied to the same dataset. In addition, the method presents the advantage of directly providing estimations of the uncertainty of the calibrated measurements, which is an important metric used to assess the quality and reliability of data provided by air pollution sensors and that most other calibration methods usually cannot provide
Reference:
TAIRA, Gustavo R.; LEAL, Adriano Galindo; SANTOS, Alessandro Santiago; PARK, Song W. Bayesian neural network-based calibration for urbana ir quality sensor. In: EUROPEAN SYMPOSIUM ON COMPUTER AIDED PROCESS ENGINEERING, ESCAPE 32, 2022, Toulouse , France. Proceedings… 6 p.
Document with restricted access. Access to the work in PDF, only for users registered in the library. Request the password at the DAIT Library Service:
https://escriba.ipt.br/pdf_restrito/178092.pdf