Runoff Prediction Method Based on Adaptive Elman Neural Network

Abstract

The prediction of medium-and long-term runoff is of great significance to the comprehensive utilization of water resources. Building an adaptive data-driven runoff prediction model by automatic identification of multivariate time series change in runoff forecasting and identifying its influence degree is an attractive and intricate task. At present, the commonly used screening factor method is correlational analysis; others offer multi-collinearity. If these factors are directly input into the model, the parameters of the model tend to increase, and the excessive redundancy and noise adversely affects the prediction results of the model. On the basis of previous studies on medium-and long-term runoff prediction methods, this paper proposes an Elman Neural Network (ENN) adaptive runoff prediction method based on normalized mutual information (NMI) and kernel principal component analysis (KPCA). In this method, the features of the screening factors are extracted automatically by using the mutual information automatic screening factor, and then input into the Elman Neural Network for training. With less features, the parameters of the Elman Neural Network model can be reduced, and the problem of overfitting of the Elman Neural Network model is effectively alleviated. The method is evaluated by using the annual average runoff data of Jinping hydropower station in Chengdu, China, from 2007 to 2011. The maximum relative error of multiple forecasts was found to be less than 16%, and forecast effect was good. The accuracy of prediction is further improved by averaging the results of multiple forecasts. methods used to establish mathematical relations include linear regression, stepwise regression [7], local regression, artificial neural networks [8][9][10], and support vector machines [11][12][13]. Meanwhile, a process-driven model requires a hydrological model that can reflect the characteristics of runoff, and future medium-and long-term rainfall information is used as model input to obtain changes in the forecast object [14]. The ensemble streamflow prediction (ESP) method proposed by American scholar Day [15] is a process-driven model and researchers have used this method to study medium-and long-term runoff forecasting in many watersheds. As the mechanism of hydrological process has not been fully elucidated, the applicability of this model is limited [16][17][18][19][20]. Therefore, a data-driven model, especially the runoff prediction model based on neural networks, has become a focused topic for [21][22][23][24] the application of back propagation (BP) neural networks to medium-and long-term hydrological forecasting [25]. In [26][27][28], the application of wavelet neural networks to runoff forecasting was investigated. In [29], the application of gray self-memory based on a BP network model to runoff forecasting was examined. However, these neural network models have two drawbacks: easy fall into local minima and slow convergence [30]. further evaluate and compare the performance of ENN and land surface hydrolo...

Annual Runoff Forecasting Based on Multi-Model Information Fusion and Residual Error Correction in the Ganjiang River Basin

“…The Elman neural network (ENN) was firstly proposed by Elman (1990) to address the voice processing problem, and it is a typical dynamic recursive neural network. Based on the basic structure of BPNN, a context layer is added from the hidden layer to the input layer in the structure of ENN, and this context layer is taken as a one-step delay operator to record information from the last network iteration as input to the current iteration [43]. In addition, ENN enables the system to adapt to time-varying characteristics, as it enhances the global stability and has stronger computing power.…”

Runoff forecasting model based on CEEMD and combination model: a case study in the Manasi River, China

“…As a forecasting model with universal approximation ability, many runoff forecasting models based on neural network have been proposed in recent years. These forecasting models include extreme learning machine (Niu et al 2018;Cheng et al 2020), RBF neural network (Wu 2018), fuzzy neural network (Shi et al 2016), and Elman neural network (Li et al 2019). Although the performance of traditional neural network is excellent, it is difficult to determine the structure of the network.…”

Levenberg-Marquardt Training Function using on MLP, RNN and Elman Neural Network to Optimize Hourly Forecasting in Tetouan City (Northern Morocco)

“…The second class of networks is recurrent networks (RNN). This is a type of loop network, for which we have chosen closed loop architecture for NARX [19] and Elman neural network [20]. NARX feedback algorithm is similar to a multi-layer Perceptron only it relies on the system's input and output regressors to train the network from which it reports a new input from the output.…”