Elemental compositions of coal are essential for analyzing the overall process of energy conversion systems

it is vital for power plant operators or process control system to be simultaneously acquaint of properties of coal and take some adjustments to abate negative impacts due to variations in coal properties. A novel method was developed to predict elements of coal using proximate analysis. Firstly

correlations of compositional content of coal were established

and residuals of the correlations were regarded as random process and predicted according to Gaussian process regression model. The ultimate analysis was achieved by solving a set of simultaneous equations of compositional content of coal and complete component integrity equation. This method can not only ensure the integrity of the whole component

but also improve the accuracy of the component prediction. The mean relative errors δ_(MAPE) of C

H

O and N were 1.46%

6.34%

16.48% and 12.86%

respectively. For coals with high sulfur content

the prediction errors of S content were high

but for coals with common S content

the prediction error was significantly lower.