Verifikace a analýza dat z čidel PM10 a PM2.5

Abstract

With the possibilities of acquiring low-cost sensors measurements of the air pollution at more affordable prices, there are also questions if these measurements are trustworthy. As a part of this thesis, it was possible to find out how low-cost Perfect-Air sensors measure the PM10 and PM2,5 particulate air pollution and whether it is possible to improve the accuracy of measurements by additional correction of measured data. For the accuracy research and refinement of the measured data, there were used several methods: correlation analysis, correction using meteorological indicators and correction using neural networks. It has been demonstrated that correction using neural networks can refine the measurement results. On the contrary, correction using meteorological indicators was not sufficient for our measurements. After fine-tuning the low-cost sensors and possibly adjusting the measured values, these devices could be used in large quantities for much more detailed monitoring of air quality.