91 / 2023-05-24 08:55:25

Mining subsidence monitoring model based on VMD-GRU based on SSA method optimization

VMD;GRU; SBAS-InSAR; Subsidence prediction

Early detection of geological risks such ground subsidence and cave-ins in mining locations can be achieved by predicting them, and prompt corrective action which is crucial for Safe mining of coal. The paper proposes a Combined forecasting method， variational mode decomposition (VMD) and a network of gated recurrent units (GRU) which is optimized using the Sparrow Search Algorithm (SSA) for the prediction of mine subsidence to address the volatility and non-smoothness of mine subsidence sequences. An intelligent optimisation technique is utilized to optimize the crucial parameters in both algorithms in order to solve the issue of difficult tuning of VMD and GRU parameters. Firstly, the optimized VMD decomposes the time series subsidence data into a series of intrinsic mode function (IMF), and make the sample entropy (SE) is used as a basis for dividing the eigenmodal components of different frequencies into two subseries of similar complexity, the GRU prediction model was then built separately to make fast and accurate predictions, and finally the prediction results were combined and superimposed to obtain the final prediction results.The SBAS-InSAR (Small Baseline Subsets Interferometric Synthetic Aperture Radar, SBAS-InSAR) technique was employed to obtain ground subsidence information for the mining area from January 2019 to April 2021 using ground subsidence in a mining area in Jinan, Shandong Province as the research object.The total mine area formed a clear funnel shape on April 5, 2021, and the maximum cumulative subsidence of the subsidence center had reached 76 cm , which indicates that the SBAS-InSAR technique can obviously monitor the subsidence trend of the study area, and the reliability of the SBAS-InSAR results was analysed and verified Using level survey data from the same period. In light of this, a combined subsidence prediction model was built using the obtained subsidence time sequences from SBAS-InSAR technique, the correlation coefficients between the test set's deformation data and the predicted data were all calculated to be 0.99,this demonstrating the effectiveness of the combined prediction model that was experimentally trained in this paper. Then the RMSE, MAE, and MAPE of this model were compared with those of other prediction models to further highlight its superiority. The results revealed that the SSA-optimised VMD-GRU model can more effectively strike a balance between operational efficiency and prediction accuracy while avoiding the tediousness and randomness of manual parameter adjustment. It offers a unique approach for the prediction of mine subsidence sequences and has great self-adaptability and intelligence.