On the basis of hot-state tests under many operating conditions for investigating the characters of unburned carbon in fly ash of boilers for mixedly burning blended coal

a neural network model has been established by using the nonlinear dynamics and self-learning characters of the artificial neural network.The relative error between predicted value and measured value are 1.4%

1.19%

and 2.45% respectively

comfirming the feasibility of said model.Taking the adjustment parameters in boiler operation as independent variables of the target function for optimization

and combining with the genetic algorithm in overall situations

directing against three typical operating conditions

namely blending Indonisian coal in 0

1

and 2 mills

the optimal operating parameters have been found and obtained

the unburned carbon content in fly ash has decreased from original 2.25%

1.96%

and 1.08% to 1.87%

1.69%

and 0.73% respectively.It shows that the mixedly burbing of Indonisian coal is helpful to decrease the unburned carbon content in fly ash

enhancing the boiler efficiency.