During the modeling of several soft sensing methods of carbon content in fly ash

the input features are not filtered by feature selection method and the modeling method type is relatively simple. To solve these problems

this article proposes a prediction model of carbon content in fly ash based on variable selection and flower pollination algorithm-optimized random forest(FPA-RF). In this model

the factors influencing the carbon content of fly ash constitute the initial variables

and the optimal variables are selected from the initial variables by the random forest-based variable selection method. Then

the random forest is trained by the optimal variables and its parameters are optimized by the flower pollination algorithm. Taken the running data of a 600 MW coal-fired power plant as the example

the prediction effects of the above FPA-RF model and the model based on artificial neural network(ANN) and particle swarm optimization-support vector machine(PSO-SVM) were compared. The results show that

the ANN prediction model has the largest prediction error and weaker generalization ability; the FPARF and PSO-SVM model both achieve better prediction effects

but the average training time of the PSO-SVM model is 55 s

which is about 30 s longer than that of the FPA-RF model. The FPA-RF model has higher prediction accuracy and faster calculation speed.