Water yield of mine analysis and prediction method based on EEMD-PSO- ELM-LSTM model

Abstract

Abstract In view of the complexity of mine water inflow data analysis and the uncertainty of prediction and prediction and other key issues, according to the data characteristics of metal mine water inflow, a method of mine water inflow analysis and prediction based on EEMD PSO-ELM-LSTM is proposed by applying the phase space reconstruction idea and the fusion modeling concept. Taking the monthly average water inflow of JIAOJIA Gold Mine in China from January 2014 to October 2021 as an example. Firstly, the Ensemble Empirical Mode Decomposition (EEMD) is used to decompose the measured data series of mine water inflow into trend components, seasonal components, and remainder components, and the remainder components are treated as noise and removed; Subsequently, based on the data characteristics of the decomposed component data, the PSO-ELM algorithm is selected to analyze and predict the seasonal components of water inflow, and the LSTM model is applied to analyze and predict the trend components of water inflow; Finally, the analysis and prediction results of the two are superimposed and reconstructed to obtain the final analysis and prediction results. In addition, comparative predictions were made using EEMD PSO-ELM-LSTM, LSTM, and EEMD LSTM. Compared with the independent prediction models LSTM and EEMD LSTM, the Root Mean Square Error (RMSE) of the EEMD PSO-ELM-LSTM algorithm proposed in this paper has been reduced by 248.04 and 76.27, respectively; Mean Square Error (MSE) decreased by 0.047 and 0.011, respectively; At the same time, the Nash-Sutcliffe efficiency coefficient (NSE) of the model proposed in this article is closer to 1. In summary, the EEMD PSO-ELM-LSTM mine water inflow analysis and prediction method has certain reliability and superiority, which helps to promote accurate prediction of average mine water inflow and reduce the occurrence of water inrush accidents in metal mines.