Author(s): Zhang Di; Peng Qidong; Wang Dongsheng; Lin Junqiang

Linked Author(s):

Keywords: No Keywords

Abstract: Reservoir operation is an important measure to realize the comprehensive benefit of the reservoir and to mitigate the adverse environmental impacts of water conservancy projects. It is crucial to full play the role of reservoir operation that develop a reasonable and effective reservoir operating plan. To explore the application of deep learning algorithm on the field of reservoir operations, Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU) are employed, and compared with respect to their capabilities for predicting outflows for Gezhouba (GZB) reservoir in China. Results show (1) a reasonable increase in the number of maximum iterations is helpful to improve the accuracy of the model, and the effect of hidden nodes on model precision is weak, the batch sizes mainly affects the calculation speed of the model, and the larger the batch sizes, the faster the calculation speed; (2) three models are capable of providing reservoir outflows with satisfactory statistics, and the applicability of recurrent neural network model in reservoir operation is verified; (3) comparing the three models, the results obtained by RNN and GRU have the optimal statistical performances compared with LSTM.

DOI:

Year: 2018