Prediction of inlet SO2 concentration of wet flue gas desulfurization (WFGD) by operation parameters of coal-fired boiler.

Abstract

Circulating fluidized bed (CFB) boilers with wet flue gas desulfurization (WFGD) system is a popular technology for SO2 removal in the coal-fired thermal power plant. However, the long response time of continues emission monitoring system (CEMS) and the hardness of continuously monitoring the coal properties leads to the difficulties for controlling WFGD. It is important to build a model that is adaptable to the fluctuation of load and coal properties, which can obtain the SO2 concentration ahead CEMS, without relying on coal properties. In this paper, a prediction model of inlet SO2 concentration of WFGD considering the delay between the features and target based on long-short term memory (LSTM) network with auto regression feature is established. The SO2 concentration can be obtained 90s earlier than CEMS. The model shows good adaptability to the fluctuation of SO2 concentration and coal properties. The root-mean-squared error (RMSE) and R squared (R2) of the model are 30.11mg/m3 and 0.986, respectively. Meanwhile, a real-time prediction system is built on the 220 t/h unit. A field test for long-term operation has been conducted. The prediction system is able to continuously and accurately predict the inlet SO2 concentration of the WFGD, which can provide the operators with an accurate reference for the control of WFGD.