In order to meet the requirement of realizing ultra-low emission of NO_x from large-scale circulating fluidized bed(CFB) units at under ultra-low load after the renewable energy connected to the power grid

the BP neural network algorithm is applied to establish the model for predicting NO_x emission mass concentration

and genetic algorithm is used to optimize the weight and threshold of the BP neural network. The effects of nine operational parameters

including boiler load

feed coal quantity

primary air

secondary air

bed temperature

flue temperature at inlet of the cyclone separator

oxygen volume fraction at the economizer outlet

main pipe flow of urea solution

and furnace bed average pressure

on the NO_x emission mass concentration of a 350 MW unit CFB boiler

are comprehensively considered

and the major optimal parameters of the unit at ultra-low load(100 MW)are obtained

which are: the main pipe flow of 20%(mass fraction) urea solution is 0.5 m~3/h

the ratio of primary air volume to secondary air volume is 2.8

the bed temperature of the CFB boiler is 855 ℃

and the flue temperature at inlet of the cyclone separator is 778 ℃.