Abstract: In order to effectively and accurately predict the absolute gas emission of the mining face, a VMD-DE-RVM interval prediction method was worked out for absolute gas emission from a mining face based on VMD method, DE algorithm and RVM principle. For working out the method, the absolute gas emission was decomposed into several intrinsic mode components by VMD method and the local characteristics were analyzed. RVM prediction models were built for the intrinsic mode component, and the model parameters were optimized by DE algorithm to improve the prediction accuracy. Prediction results of the components were weighted and superimposed to obtain an absolute gas emission prediction result. A prediction result comparison was made between the three models. The results show that the method worked out by the authors can weaken the local characteristics of gas emission, and the prediction interval effectiveness is 100%, the average absolute error is 0.096 m³/min and the average relative error is 2.43% when the confidence is 95%, and the prediction accuracy is improved.

Key words: absolute gas emission, interval prediction, variational mode decomposition(VMD), relevance vector machine(RVM), differential evolution(DE) algorithm