Development of a multi criteria decision support system using convolutional neural network and jaya algorithm for water resources management / Chong Kai Lun
This dissection aims to develop and deploy a multicriteria support system framework to provide a structured decision-making process. The proposed approach can be furthered categorized into two distinct stages: forecasting modeling and optimization modeling. Artificial neural network (ANN) has been w...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/13813/1/Chong_Kai_Lun.pdf http://studentsrepo.um.edu.my/13813/2/Chong_Kai_Lun.pdf http://studentsrepo.um.edu.my/13813/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This dissection aims to develop and deploy a multicriteria support system framework to provide a structured decision-making process. The proposed approach can be furthered categorized into two distinct stages: forecasting modeling and optimization modeling. Artificial neural network (ANN) has been widely used in forecasting tasks. However, due to some drawbacks, an advanced technique was employed in this study. The proposed method involves using a convolutional neural network (CNN) with a feature extraction ability to learn from the hydrological dataset efficiently. With the aid of deep architecture, a highly abstracted representation of the inputs time series with a high level of interpretation is formed at each subsequent CNN layer. Besides, through integration of wavelet transform (WT), the performance of the forecasting model can be improved. WT can be used to preprocessing the hydrological dataset into a set of decomposed wavelet components. These components are served inputs for the CNN model. The developed models were applied to three different case studies to evaluate the performance of the models. The results showed that the proposed model could capture patterns of the monthly and daily interval of the hydrological time series. Apart from that, having low values in four of the performance criteria: RMSE, MAE, NSE, and RSR, have further strengthened the credibility of the results. As for the optimization process, the reservoir operation rule was derived using a meta-heuristic algorithm at the monthly interval. These operational rules were based on the reservoir with a multi-purpose objective: hydropower and intrusion of saltwater. The results indicated that the hydropower generated by the proposed algorithm could produce an evenly distributed high amount of energy increases the reliability of the reservoir system. However, under the circumstances of water deficiency, the hydropower output is significantly reduced. When deriving the optimal operating rule, a hedging rule was applied to attenuate the effect of limited water supply. Furthermore, the efficiency of the proposed algorithm was assessed using some reservoir performance indices such as resilience and reliability. Besides, a Bayesian uncertainty analysis was carried out to quantify the model output behaviors due to derivation from the uncertainty in the input parameters. A Bayesian method for CNN using TensorFlow Probability was used in this study. By utilizing the probabilistic model, the aleatoric and epistemic uncertainty can be addressed. In addition, the confidence level was built using the percentile-t-method (or bootstrap-t-method). The proposed technique was then tested on a dataset obtained from the same hydrological stations used when the forecasting modeling. According to the simulated results, the proposed model can provide a statistical distribution of the forecasted quantity. Besides, the Monte-Carlo simulations demonstrated that all the values lie within the 95% confidence level. Therefore, the network reliability increased as it revealed the uncertainty in the forecasted values.
|
|---|