Kinetic characteristics of the particles in a dense-phase pulsed fluidized bed for dry beneficiation
点击次数:
关键字:As a clean energy source, coalbed methane has broad prospects in the global energy structure. Due to the strong heterogeneity of coalbed methane production, it is of great practical significance to obtain the key control factors of coalbed methane production types. Accurate prediction of its output is important, but it is a challenge to accurately predict the output of coalbed methane because it is affected by a variety of factors. In previous studies, many researchers have used deep learning algorithms to predict the output of coalbed methane, which are mainly based on image or sequence data, and rarely consider the nonlinear spatial coupling relationship between coalbed methane wells. In order to solve the above problems, an interpretable residual graph convolutional neural network model (I-RGCN) was proposed for the classification of coalbed methane production types. Based on the prior geological knowledge, the topological map structure is constructed, and the geological similarity between coalbed methane wells is measured by the DTW algorithm and added to the model calculation as the edge weight of the graph structure, and finally the Gnnexplainer is used to rank the feature importance of the model prediction. Experiments on the dataset in the Fanzhuang-Zhengzhuang block of Qinshui coalfield in Shanxi Province show that the interpretable residual map convolutional neural network model has an accuracy of 84%, an accuracy of 68%, and a recall rate of 61%, and the comprehensive index score is higher than that of other baseline comparison models. This study provides a new explanatory prediction method for the prediction of coalbed methane production type, and also provides a new way to use graph neural networks to deal with similar coalbed methane production problems.
是否译文:否