Abstract: The prediction of ground subsidence and deformation in coal mines mainly based on the theory of subsidence prediction of coal mining. Deformation prediction models based on deformation analysis theory currently focus on single model prediction. Based on the combination forecasting idea, a non-equidistant gray prediction model and a BP neural network model are used as prediction single models. The measured surface settlement value above the coal mining face of a coal mine in northern Shaanxi is used as the data source. The optimal weight combination is used to optimize the prediction results of the single model, and the weights of the two single models in the combined model are 0.466 7 and 0.533 3. The prediction results of some monitoring points are selected to evaluate the model accuracy, and the results show that the accuracy of all three prediction models reached one level. By comparing the prediction results of the three models, the accuracy of the optimal weight combination prediction model is significantly improved compared with the single model, and the prediction results have significant gains compared with the non-equidistant grey prediction model and the BP neural network prediction model.

Key words: subsidence deformation prediction, grey system model, BP neural network model, combination forecast, optimal weight combination