Characteristic Analysis and Prediction of Runoff Based on Chaotic Wavelet Neural Network

This paper proposes a method combining phase space reconstruction and chaotic wavelet neural network to predict runoff. The mutual information method and Cao method are selected to calculate the delay time and embedding dimension, and the Lyapunov exponent is used to judge the chaos. In the chaotic wavelet neural network model, the reconstructed time series is used as the input of the neural network, and the wavelet function is used as the activation function of the neuron. Applying the model to the prediction of the daily runoff and monthly runoff of the Mississippi River, the results show that the model is effective for predicting data of different scales and still has high precision under a long prediction period.