A calibration methodology of low-cost air pollutant sensor using neural networks

Aymane Souani (1) , Alexandre Hucher (1) , Vincent Vigneron (2) , Hichem Maaref (2)

1 Ecomesure
2 IBISC - Informatique, BioInformatique, Systèmes Complexes

Air quality Low cost sensors (LCSs) are cheap and can map extensive areas. They alert people about pollution spikes in smart city buildings (schools, universities, hospitals. . .) or industrial areas. Before using them for a specified task, they must be calibrated to give accurate readings, i.e. they must be aligned with a measure based on a reference machine. Unfortunately, classic calibration is limited by interferences with other pollutants or can be affected by atmosphere constants in the case of uncontrolled environments. This paper proposes a calibration solution based on artificial neural networks (ANN).

Keywords: Calibration; Air pollutant; Regression; Neural netwoks; SVR; LCS

Domains: Signal and Image processing; Neural and Evolutionary Computing [cs.NE]; Environmental Sciences

Aymane Souani, Alexandre Hucher, Vincent Vigneron, Hichem Maaref. A calibration methodology of low-cost air pollutant sensor using neural networks. 20th IEEE International Multi-Conference on Systems, Signals & Devices (SSD 2023), Feb 2023, Mahdia, Tunisia. pp.229--235, ⟨10.1109/SSD58187.2023.10411311⟩. ⟨hal-04367082⟩