关键词: 露天矿　SBAS-InSAR 方法　形变预测　PSO-LSTM 模型　粒子群优化算法　长短期记忆模型

Abstract: Analyzing and predicting the characteristics and trends of the surface deformation of open-pit mines is an important
part of ensuring the safe and green operation of mines. For large-scale open-pit mines,taking the Junning Gobi No. 2
open-pit mine in Xinjiang as an example,based on the SBAS-InSAR method and the particle swarm optimization algorithm′s
long short-term memory network (PSO-LSTM) model,a method for analyzing and predicting the surface deformation of openpit
mines is proposed. This method first calculates the surface deformation of the mine using the SBAS-InSAR method,and
then,in response to the problems such as low efficiency and limited spatial coverage of current deformation monitoring methods
like leveling measurement and GNSS in large-scale open-pit mines,the particle swarm optimization algorithm (PSO) was used
to optimize the long short-term memory model (LSTM),and a PSO-LSTM model was constructed for deformation prediction.
The research shows that:① The overall average deformation rate of the mining area is -2. 832 mm/ a,showing a downward
trend. The surface deformation rate of the inner waste dump area is significantly higher than other areas;spatially,the inner
waste dump area and the east waste dump area are distributed relatively evenly;temporally,the deformation rates of the east
and north waste dump areas are lower,and the rates are relatively constant. ② Through the profile lines,it can be found that
the spatial deformation distribution of the north waste dump area shows non-uniformity,while the east waste dump area exhibits
relatively balanced deformation characteristics. The Root Mean Square Error (RMSE),Mean Absolute Error (MAE),Mean
Absolute Percentage Error (MAPE) and coefficient of determination (R2) are adopted as the evaluation indicators for the prediction
accuracy. The results show that compared with support vector regression (SVR) model and LSTM model,the RMSE and
MAE of the PSO-LSTM model are at least reduced by 16% and 30%,respectively. The PSO-LSTM model has better stability
and smaller deviation,reflecting that this model can effectively capture the fluctuation trend of the surface deformation of the
mining area and has certain stability. The research results provide new ideas for the analysis and early warning of surface deformation
of open-pit mines and have certain reference significance for the monitoring and prediction of surface deformation of
large-scale open-pit mines.

Key words: open-pit mine,SBAS-InSAR method,deformation prediction,PSO-LSTM model,genetic algorithm optimization,
long short-term memory