Circulating fluidized bed boiler state estimation with an unscented Kalman filter tool

The paper discusses the development of a state estimation tool for circulating fluidized bed (CFB) boiler dynamic hotloop models. Bayesian state estimation was used to determine inputs, states and time-variant parameters based on output observations. The goal was to apply advanced state estimation to the original nonlinear model and utilize it for reallife CFB applications. The main algorithm of the tool was the unscented Kalman filter (UKF), with an SIR particle filter as a backup solution. The implementation of the tool and the UKF algorithm were described. The tool was tested with two simulation cases. In the first case, fuel flows and an air leakage parameter were identified based on flue gas compositions for pilot oxy combustion measurements. In the second case, heat transfer coefficient and fuel moisture content values were estimated in an industrial boiler based on the dense bed furnace temperature and the flue gas O₂ content. The results showed a good agreement between measurements and simulations, as well as a good computational performance for the UKF.