Abstract:

 In order to predict the ash fusion point of a mixed biomass more quickly and accurately, the support vector machine(SVM)regression model was optimized by a genetic algorithm(GA), built by other researchers was further optimized by cross validation(CV). The ash fusion point of a mixed biomass was predicted by the optimized model and was   trained by the data of a single biomass while  taking ash compositions as input and the ash fusion point as output. The result was compared with models optimized only by GA. The results showed that the SVM model optimized by GA and CV, with average absolute error 25。0℃ and relative error 2。7％, could predict the ash fusion point of a mixed biomass better than that  optimized only by GA, and the running time could be saved if parameters were  properly set.

Key words: ash fusion point, support vector machine, genetic algorithm, cross validation, prediction