A Flame Imaging-Based Online Deep Learning Model for Predicting NO? Emissions From an Oxy-Biomass Combustion Process

To reduce NO x (nitrogen oxide) emissions from fossil fuel and biomass-fired power plants, online prediction of is important in the plant operation. Data-driven models have been developed to predict various combustion processes with good accuracy. However, such initially built based on known conditions, which are historically “seen”. For new “unseen”, these usually perform unwell. In this study, an deep learning (ODL) model proposed oxy-biomass process for “seen” “unseen” conditions source condition recognition models. The ODL mainly conditions. A objective function that consists regression loss distillation introduced improve examined using boiler operation data, flame temperature maps, data obtained under a range Oxy-Fuel Combustion Test Facility. Flame images acquired dedicated imaging system used computing distribution through two-color pyrometry. results demonstrate capable predicting mean absolute percentage error less than 3%, first, second, third updates.