Developing a robust and scalable calibration approach to low-cost AQ sensing

TECHNICAL ABSTRACT: Poor air quality represents a major public health risk, contributing to an estimated one of every eight deaths worldwide. Low-cost air quality sensors have developed rapidly over the last few years and offer the opportunity to make air quality monitoring inexpensive and widespread. However, low-cost sensors are currently unable to deliver accurate, reliable data due to a lack of understanding of what parameters impact performance and due to calibration models that inadequately describe sensor behavior under ambient conditions. This proposal will develop improved calibration models for electrochemical sensors utilizing realistically constrained, shortterm laboratory calibration chamber experiments and long-term (24-month) sensor-to-reference co-location field deployments. Various machine learning techniques will be trained and evaluated using sensor and reference data from both calibration chamber experiments and ambient co-located field data. The end result of the Phase I project will be a robust calibration method that clearly identifies the parameters that impact performance over the lifetime of electrochemical sensors. The Phase I effort will provide the empirical foundation from which the optimized machine learning techniques can be adapted in Phase II to provide robust, universal calibration methods for electrochemical sensor systems at scale.SUMMARY OF ANTICIPATED RESULTS: The calibration method for low-cost air quality sensors developed in this project will lead to widely expanded measurements of air quality and to improved understanding and mitigation of air pollution. It will also increase sales of Aerodyne’s low-cost air quality sensor package.