In order to achieve the carbon emission peak and carbon neutrality goals

it is imperative to reform and upgrade the combustion system of utility boilers. Firstly

a new improved sparrow search algorithm(ISSA) is developed by using elite opposition-based learning strategy

adaptive dynamic inertia weight and adaptive t-distribution mutation

to improve the population initialization and position update of sparrow search algorithm(SSA). Then

by optimizing the regularization coefficient and kernel function parameters of kernel extreme learning machine(KELM) by ISSA

the prediction model of boiler combustion characteristics is established. The model is used to predict acutral operation data of an ultra supercritical 660 MW unit

and the average absolute error rate of the predicted NO_x emission mass concentration and boiler thermal efficiency is 1.7191% and 0.0239%

respectively

indicating the prediction effect is good. Finally

based on the prediction result

the ISSA is used again to optimize the adjustable operating parameters of the utility boiler. After optimizaiton

the NO_x emission mass concentration reduces by about 91.73 mg/m~3 at low load and 45.96 mg/m~3 at high load

and the boiler thermal efficiency increases by 0.54% at low load and 0.50% at high load.