Characteristic Analysis And Prediction Research Of Mine Water Inflow Time Series

Mine water inflow has always played a key role in coal mine safety production.Clarifying the occurrence mechanism of water inflow,mastering the variation characteristics of water inflow,and predicting the reliability of mine water inflow can provide effective guarantee for underground safe production.However,due to the influence of various factors such as mining area,development length,aquifer properties,precipitation,geological structure and other factors in the actual production of the mine hydrological system.These factors make the system as a whole have obvious complexity,nonlinearity and non-stationary variability.In addition,with the massive consumption of fossil energy,the mining depth of coal mines is gradually deepening.Therefore,the study of deep underground hydrological system can provide guidance for mine production and mining activities.From the perspective of time series,this paper took three typical North China coal mines,Wangxingzhuang Coal Mine,Xin’an Coal Mine and No.10 Coal Mine of Pingdingshan as examples,and took the water inflow in the main mining area of the2-1 Coal Seam at different burial depths in the third mining area for the research object.Chaos theory,recurrence plot theory,pearson correlation coefficient method,artificial neural network were introduced to qualitatively describe and quantitatively analyze the dynamic characteristics of the time series of mine water inflow in different mining areas and different burial depths,and construct the C-ANN prediction model of mine water inflow.The research mainly achieved the following results:（1）Phase space reconstruction of mine water inflow time series.The chaotic parameters were calculated by the mutual information method and the Cao method.And respectively for Wangxingzhuang Coal Mine from May 2006 to March 2020（167 months）,Xin’an Coal Mine from January 1984 to October 2020（442 months）,No.10 Coal Mine of Pingdingshan from January 2009 to June 2020（138 months）of the monthly water inflow time series for phase space reconstruction.The chaotic parameters of different mining areas in different time periods were calculated as follows:the delay timeτof water inflow in Wangxingzhuang Coal Mine was 4months,and the embedding dimension m was 6（In order to properly describe the variation characteristics of the measured monthly water inflow time series and build a dynamic model,6 independent variables are required,that is,6 influencing factors are required）;the delay timeτof water inflow in Xin’an Coal Mine was 6 months,and the embedding dimension m was also 6（6 influencing factors）;the delay timeτof water inflow in No.10 Coal Mine of Pingdingshan was 2 months,and the embedded dimension m was 5（5 influencing factors）.In order to continue to compare the chaotic characteristics of water inflow in the three mining areas in the horizontal direction,the time series of water inflow in the three mining areas from January 2009 to March2020（135 months）were selected for calculation.It was concluded that the delay timeτof water inflow in Wangxingzhuang Coal Mine was 3 months,and its embedding dimension m was 6（6 influencing factors）;the delay timeτof water inflow in Xin’an Coal Mine was 4 months,and its embedding dimension m was 5（5 influencing factors）;the delay timeτof water inflow in No.10 Coal Mine of Pingdingshan was 2months,and its embedded dimension m was 5（5 influencing factors）.（2）Chaos characteristic analysis of mine water inflow time series.By using the maximum Lyapunov exponent and Kolmogorov entropy,the chaos characteristics of water inflow in Wangxingzhuang Coal Mine,Xin’an Coal Mine and No.10 Coal Mine of Pingdingshan were discriminated and analyzed.First,the Wangxingzhuang Coal Mine from May 2006 to March 2020（167 months）,Xin’an Coal Mine from January1984 to October 2020（442 months）,No.10 Coal Mine of Pingdingshan from January2009 to June 2020 Monthly（138 months）monthly water inflow time series was calculated.The maximum Lyapunov exponent and Kolmogorov entropy of the obtained results in each mining area were both greater than zero,indicating that the time series of water inflow in the three mining areas had obvious chaos characteristics.Then,the time series of water inflow in the third mining area from January 2009 to March 2020（135 months）was selected for comparative analysis.It was calculated that the maximum Lyapunov exponent of the there mining areas in this period were0.0878,0.1246,0.0951;the Kolmogorov entropy was 0.3992,0.5636,0.4867 in turn.By comparison,it was found that there was a certain relationship between the burial depth of the 2-1 Coal Seam and the chaos of the mine water inflow.With the increase of the burial depth of the 2-1 Coal Seam,the chaos of water inflow showed a decreasing trend.（3）Recurrence characteristic analysis of mine water inflow time series.The thresholded recurrence plot and the non-threshold recurrence plot of the water inflow time series of Wangxingzhuang Coal Mine from May 2006 to March 2020（167months）,Xin’an Coal Mine from January 1984 to October 2020（442 months）and No.10 Coal Mine of Pingdingshan from January 2009 to June 2020（138 months）were drawn respectively.By analyzing the shape or color of the surface in the recurrence plot,the chaos characteristics of the visualized water inflow were explained;then the recurrence quantitative parameters in different mining areas and different time periods were calculated.Then,draw the recurrence plot with threshold and recurrence plot without threshold of water inflow from January 2009 to March 2020（135 months）in the third mining area.And it was calculated that the DET of the three mining areas were 0.8929,0.5556,0.6989;the L_meanof the three mining areas were 7.1920,2.5000,6.0645;the LAM of the three mining areas were 0.9388,0.7157,0.8030,respectively.Since recursion is associated with chaos in system dynamics,the recursion of water inflow is also associated with the burial depth of the 2-1 Coal Seam.（4）Establish an ANN water inflow prediction model.Through the analysis of the factors that may cause water inflow in the 2-1 Coal Seam of Wangxingzhuang Coal Mine,the goaf area（S）,development length（L）,precipitation（P）,and buried depth（H）,which mainly affected the water level of the aquifer,were determined as the main influencing factors.Among them,the water source factors were C₂t L_7-8limestone aquifer,C₂t L_1-4limestone aquifer,O₂m limestone aquifer,∈₃ch limestone aquifer.The above variables were quantified as time series statistical values and used as input elements of the input layer of the artificial neural network model to train and build the network to predict the water inflow.After calculation,the prediction accuracy of BPNN model was 93.65%,and the prediction accuracy of ENN model was 95.50%.（5）Combining chaos theory with artificial neural network to build a C-ANN water inflow prediction model.Taking Wangxingzhuang Coal Mine as the main research object,through in-depth discussion of the relationship between each dimension of the reconstructed water inflow time series and its actual influencing factors,the mechanism of the specific C-ANN water inflow prediction model was clarified.The model was tested in Xin’an Coal Mine and No.10 Coal Mine of Pingdingshan.The results showed that the C-ANN prediction model can play the respective advantages of chaos theory and artificial neural network,and had certain universality and reliability.In Wangxingzhuang Coal Mine,the prediction accuracy of C-BPNN was 91.35%,and the prediction accuracy of C-ENN was 95.71%;in Xin’an Coal Mine,the prediction accuracy of C-BPNN model was 92.45%,and the prediction accuracy of C-ENN model was 94.78%%;in No.10 Coal Mine of Pingdingshan,the model accuracies of C-BPNN and C-ENN were 89.8%and 93.26%,respectively.