Application of Recurrent Neural Network for Inflow Prediction into Multi-purpose Dam Basin

Abstract

This paper aims to investigate the applicability of dam inflow prediction model using recurrent neural network theory. To achieve this goal, the Artificial Neural Network (ANN) model and the Elman Recurrent Neural Network (RNN) model were established for Soyanggang and Chungju dam basin using meteorological and hydrological data accumulated from dam operation since constructed. For the model training, inflow, rainfall, temperature, sunshine duration, wind speed are used as input data and daily inflow of dam for 10 days are applied as output of the model. And predictions of dam inflow for 2 years from July, 2016 to June, 2018 are carried out for verification purpose. The results showed that there was no significant difference in prediction performance between ANN model and the Elman RNN model in Soyanggang dam basin but the prediction results of the Elman RNN model are comparatively superior to those of the ANN model in Chungju dam basin. Consequently the Elman RNN prediction performance is expected to be similar to or better than the ANN model. Especially, the prediction performance of Elman RNN is also superior during low dam inflow period. In addition, it is shown that the multiple hidden layer structure of Elman RNN is analyzed to be more effective in prediction performance improvement than single hidden layer structure.