Development of operation multi-objective model of dam reservoir under conditions of temperature variation and loading using NSGA-II and DANN models: a case study of Karaj/Amir Kabir dam

• Published in: Soft computing (Berlin, Germany) Germany), 2020-08, Vol.24 (16), p.12469-12499
• Main Authors: Tabari, Mahmoud Mohammad Rezapour, Azadani, Mitra Nasr, Kamgar, Reza
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Artificial neural networks; Back propagation; Computational Intelligence; Computer simulation; Control; Dam stability; Dams; Decision making; Dynamic programming; Dynamic stability; Engineering; Finite element method; Fuzzy logic; Genetic algorithms; Human error; Hydroelectric power; Irrigation; Mathematical Logic and Foundations; Mechatronics; Methodologies and Application; Methods; Multiple criterion; Multiple objective analysis; Neural networks; Optimization; Parameters; Reservoirs; Robotics; Seasonal variations; Simulation; Simulation models; Sorting algorithms; Stability analysis; Structural stability; Summer; Support vector machines; Water levels; Water shortages; Water supply
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-020-04686-1

For determining optimal operation policies, it is of vital importance to the structural stability of dams, besides meeting the downstream water demands during the operation period. Since water level variations in upstream dam cause changes in loadings of the dam body, ignoring factors, which are effective in the structural stability of dams, might result in cracks and instability of dams in the long term. Therefore, in the present study, a multi-objective model was developed by integrating the reservoir structure simulation model and optimization approach in order to meet water supply demands and maintain the dam structural stability. Linear static analysis of Amir Kabir dam was performed using ABAQUS 6.14.3 software to extract the structural parameters, indicating the dam structural stability. For this purpose, a dynamic artificial neural network was also used as an interfaced model to relate the results of the simulator model to those of the proposed approach. The best solution was extracted from the optimal trade-off curve, which was developed using non-dominated sorting genetic algorithm (NSGA-II) and multi-criteria decision-making methods. Results showed that the allocation to the downstream water demands was applied in the best possible way based on the stability of the dam. A comparison between the existing and optimal conditions indicated that the value of reliability and vulnerability coefficients was improved in optimal conditions in comparison with the existing conditions. According to the results, there was a 7% rise in the fuzzy stability index in optimal conditions compared to the existing conditions that indicate better performance in the optimal model. Therefore, as to consider the downstream water demands, summer was the most deficient and spring was the least deficient with the water supply of 30% and 72%, respectively. During summer, the average optimal allocation and the average demand were 20.51 MCM and 67.50 MCM, respectively. However, this season showed the best performance in preserving the dam stability with having the lowest parameters of the dam structural stability compared to other seasons. Finally, the support vector regression method was used to develop an optimal operational policy based on optimal allocation values obtained from the proposed model structure. The optimal developed policy can be used to determine optimal allocation based on hydrological parameters of the dam and its water demands. Meanwhile, it can maintain d