A machine learning field calibration method for improving the performance of low-cost particle sensors

Measurements of airborne particles in buildings with low-cost optical particle counters (OPCs) are often inaccurate and subject to uncertainties. This study introduces a methodology improve the performance OPCs measuring indoor through machine learning. A two-month field measurement campaign was conducted an occupied net-zero energy house. The studied (OPC–N2, Alphasense Ltd.) report size fractionated concentrations from 0.38 17.5 ?m. Co-located reference instrumentation included scanning mobility sizer (SMPS: 0.01–0.30 ?m) (OPS: 0.30–10 ?m). learning calibration method applies Gaussian Process Regression (GPR) includes two components: (1.) correction size-resolved OPC counting efficiency 10 ?m (2.) prediction volume distributions (mass proxy) below detection limit OPC. is applicable that concentrations. In (1.), GPR function used correct between using OPS as reference. (2.), second predict distribution SMPS/OPS done given significant contribution sub-0.38 accumulation mode. resulted improvement accuracy size-integrated (PV2.5, PV10) reported by compared SMPS/OPS. Improvements were seen Pearson coefficient (before correction: 0.59–0.83; after 0.98–0.99); determination 0.35–0.69; 0.97–0.98); mean absolute percentage error 35–69%; 19–25%).