Predicting indoor PM2.5/PM10 concentrations using simplified neural network models

Neural network models were presented for prediction of indoor concentrations of particulate matters (PMs). Indoor PM concentrations are generally determined by the outdoor concentration, the indoor generation rate, and the air change rate of a building. In this study, indoor PM2.5 and PM10 concentrations were modelled in a single office room installed with a portable air purifier (AP) and a heat recovery ventilator (HRV), using three different neural network models with various input variables. The relative importance of individual input variables indicated that as opposed to PM10, PM2.5 is more affected by outdoor origin than indoor source. This is generally consistent with previous findings explaining that the main source of PM2.5 is outdoor environment, whereas that for PM10 is indoor human activities. The simplified models can be easily applied in practice based on the CO2 concentration measured in a room and the outdoor PM concentration data acquired from public data.