Neural Network Model Coupled with Phase Space Reconstruction and Its Application

The mine water system is a nonlinear and dynamic system influenced by such factors as hydrological and geological conditions. Although neural network model can effectively solve the nonlinear problems, it´s difficult to obtain the satisfying results by direct prediction for the time series of water yield of mine because of chaotic characteristics in the system. In order to discover the nonlinear structure and movement law of the complicated system, chaos theory was introduced. In this paper, the forecasting model for water yield of mine was put forward by combination of the phase space reconstruction and neural network. One-dimension time series of water yield of mine was developed to multi-dimension series by means of phase space reconstruction, and multi-dimension series included the ergodic information, so more rich information can be excavated for neural network training. Then, on the basis of the embedding dimension of the time series, the structure form of neutral network was constructed. Finally, as an example of water yield of mine of the eighth mine of Hebi city, the model was built to forecast quantity of gushing water in mine pit, and the comparison of the network with AR(2) model was given. The result shows the model has good forecasting precision.