To avoid the model complexity and over-fitting problems caused by over large feature variable dimension during intelligent algorithm modeling of the boiler combustion system

the principal components are extracted by principal component analysis(PCA) algorithm

and the model parameters are optimized by particle swarm optimization(PSO) algorithm

thereby the PCA-PSO-LSSVM boiler efficiency prediction model is established. The research results show that

the PCA-PSO-LSSVM model has higher prediction accuracy and stronger generalization ability. The average relative error of the boiler efficiency model test set data with the largest error is only 0.002 49%

and the mean square error is 0.004 51. The average relative error of the PSO-LSSVM model test set after PCA extraction of principal components is –0.034 90%

and the mean square error is 0.019 27. The average relative error of the LSSVM model test set is 0.28%

and the mean square error is 0.459 39. It can be concluded that

the PCA-PSO-LSSVM model can accurately predict the thermal efficiency of the boiler

which has stronger adaptability

lower model complexity and a slightly higher training speed. The establishment of the PCA-PSOLSSVM model provides an important mean for multi-faceted data prediction of power station boilers.