Time series prediction of mine pressure in distributed optical fiber monitoring based on ARMA-SVR

Abstract In this study, the distributed optical fiber is used to monitor the overburden changes during the coal mining process, and the average frequency shift change degree of optical fiber is introduced as the index to judge the periodic pressure. The ARMA-SVR model composed of auto-regressive moving average (ARMA) and support vector machine regression (SVR) is established to predict the rock pressure behavior through the prediction of the frequency shift change of optical fiber. Firstly, a similar material model is built to obtain the data of the frequency shift of the optical fiber during the excavation of the working face; Then, the frequency shift change sequence of optical fiber is converted into a time series. Finally, a combined model is built. The linear auto-correlation part of the frequency shift change sequence is predicted by ARMA model, and the nonlinear residual part is predicted by SVR model to jointly complete the prediction of the frequency shift change of optical fiber. The experimental results show that the combined model proposed in this paper is superior to the single auto-regressive moving average model in different data sets, and the ARMA-SVR combined model can effectively predict the change of optical fiber frequency shift and provide a new idea for the prediction of rock pressure in the mining process.