Prediction of mine water flow based on singular spectrum analysis and multiple time-series coupled model

Abstract

To ensure the safe production of coal mine, it is significant to analyze the main influencing factors of mine water inflow, clarify the law of water inflow, and establish the prediction model of mine water inflow. In this study, a multi-means fusion prediction model is proposed to improve the prediction accuracy of water inflow by analyzing the variation characteristics of water inflow in Yunhe Coal Mine. Firstly, singular spectrum analysis technology is used to decompose the water inflow time series into three parts: trend series, fluctuation series, and noise series, and based on different information components, different models are used for fitting and prediction. Furthermore, on the basis of singular spectrum iterative interpolation, the inner and outer loop codes are written in MATLAB software, so as to establish the trend series prediction model. Additionally, according to the variation characteristics of the trend series and the noise series, the ARIMA and SARIMA models are used to fit and predict them respectively, then model structure is determined based on Bayesian information criterion (BIC), and the fitting effect of the model is evaluated by white noise test of residual series. Finally, the predicted water inflow can be obtained by linear fusion of three prediction results. By calculating the coefficient of determination (R 2 ) and the mean absolute percentage error (MAPE) of the multi-method fusion model and the single model, it is found that the R 2 and MAPE of the multi-method fusion prediction model are 94% and 2.34% respectively, which are better than the single model (74% and 6.8%). The results show that the multi-method fusion prediction model is suitable for the prediction of water inflow in new mining areas and can achieve the effect of guiding production.