In this paper

a 600 MW swirl-opposed coal-fired boiler is taken as an object to carry out research on the coking problem occurs at the bottom of the screen when the low-quality coal is burned. SEM and XRD analysis of the coke blocks were carried out and the composition of coal ash and ash melting point were tested. It pointed out that

the improper blending ratio of low ash melting point coal was the main reason for coking

and a solution was proposed from the perspective of precise coal blending and burning. Based on the data of different coal blends

the ash melting point prediction model of ACO-BP neural network was established. By comparing with the on-site measured flue gas temperature at the screen bottom

the maximum blending ratio curves of different coal blends were obtained to guide the operation adjustment work to gain the maximum economic benefits on the premise of ensuring safety. Moreover

a pre-warning software platform for utility boiler coking has been developed for timely and accurate adjustments operation

which has been used for more than a year and obtained good application results.