|
Optimized Deep Learning Procedure by Adaptive Parameters Based Genetic Algorithms for Determining Reservoir Inflow |
|---|---|
| รหัสดีโอไอ | |
| Creator | Krotsuwan Phosuwan and Panuwat Pinthong |
| Title | Optimized Deep Learning Procedure by Adaptive Parameters Based Genetic Algorithms for Determining Reservoir Inflow |
| Contributor | - |
| Publisher | TuEngr Group |
| Publication Year | 2564 |
| Journal Title | International Transaction Journal of Engineering, Management, & Applied Sciences & Technologies |
| Journal Vol. | 12 |
| Journal No. | 10 |
| Page no. | 12A10P: 1-11 |
| Keyword | Rainfall-Runoff model, Reservoir Operation, Multilayer Perceptron, ANNs, AGA, APOGA, APGA, Kaeng Krachan reservoir, Phetchaburi river basin, Optimum deep learning parameters, Remaining Life Time (RLT), Water resource manangement, Reduced flood, Reservoir management. |
| URL Website | http://TuEngr.com/Vol12_10.html |
| Website title | ITJEMAST V12(10) 2021 @ TuEngr.com |
| ISSN | 2228-9860 |
| Abstract | The determination of the reservoir inflow would be directly affected the efficiency of reservoir operation. Artificial intelligence techniques such as Artificial Neural Networks, Deep Learning (DL), and Genetic Algorithms (GAs) have been applied to many case studies of water resource management, for example, the determined relationship between rainfall and runoff, and rainfall forecast. DL has been successful for the rainfall-runoff model, but the performance of the model depends on its parameters that take more time-consuming for model development and is difficult to determine the optimum values. This paper presents the development of the Adaptive Parameters Based Genetic Algorithms (APGA) model to explore the optimum procedure of deep learning for reservoir inflow simulation for the Kaeng Krachan Reservoir and compare performance with the Adaptive Genetic Algorithm (AGA). The current study found that the mean absolute percentage error (MAPE) of the reservoir inflow from APGA was lower than AGA in all periods, so the optimum DL procedure from APGA outperforms AGA, while the DL layer architecture from APGA was more complex than AGA. In summary, APGA may be suitable for determining optimum DL procedure than AGA, but the probability of crossover (pc) and probability of mutation (pm) parameters should be studied in the future. |
