Forecast of fly-ash carbon content in a fuel-graded combustion boiler by using a local projection network

The pulverized coal combustion efficiency of a thermal power plant is reflected in the fuel burn-out degree and can be usually evaluated by way of the fly-ash carbon content of a boiler. A foreknowledge of this parameter is of the utmost importance for fuel-graded combustion optimization, i.e. for assuring a high combustion efficiency of pulverized coal simultaneously with the reduction of NO x emissions, thus enhancing boiler efficiency. Analyzed were the influencing factors of the boiler fly-ash carbon content. By using the specific features of a local projection network (LPN), such as simple structure, quick convergence speed, strong generalization ability and suitablity for a non-linear time-variation process, a model for undergoing training and the dynamic prediction of boiler fly-ash carbon content was established and tested by use of the data obtained form boiler hot-state tests. The results show that the model in question has forecast relatively accurately the carbon content of fly ash, thus providing a model basis for the fuel graded-combustion optimization.