Application of grey BP neural network model based on wavelet denoising to predict the residual settlement of goafs

Abstract

The residual settlement of goafs is a nonlinear process with time series. To study its settlement law and prediction model, we chose the Mentougou mining area in Beijing as a example. The wavelet threshold denoising method was used to optimize the measured data, and the Grey GM (1,1) and BP neural network models were combined in series. A grey BP neural network model based on wavelet denoising was proposed, the prediction accuracy of different models was calculated, and the prediction results were compared with the original data. The results showed that the prediction accuracy of the grey BP neural network combined model was higher than that of the single GM (1,1) model. Furthermore, the mean absolute percentage and root mean square errors of the grey BP neural network model after wavelet denoising were 0.35% and 16.05, respectively, both less than the errors of the combined prediction model using the original data. Thus, the combination model optimized by wavelet analysis had a high prediction accuracy, strong stability, and accorded with the law of change of measured data. It accurately reflected the goaf surface subsidence process, and therefore has a strong popularization and application value.