Abstract
The need to optimize the operation of water reservoirs is an issue that is becoming increasingly a concern for water resources planners in developing countries. This issue particularly becomes more significant in large systems with multiple reservoirs where operation of one reservoir has an impact on the others. In other word, the set of reservoirs in these systems act like a united series and require specific methods to handle various modeling issues. Problem statement: Sirvan River basin in west of Iran, standing in fifth order in respect of discharge, is an example of such a complex system. The project of water transfer from western tropical regions which is one of the large-scale projects in water resources management in Middle East, consists of a number of reservoirs and transfer systems. The matter of optimum operation of such a collection is one of the complicated and outstanding issues in water resources management. Approach: It was found that, due to increasing decision-making variables, conventional models used in optimizing water resources systems were not any longer capable of obtaining a desired solution, either because of low precision or time constraints. Hence, the intelligent random research approach "Simulated Annealing" has been used which in recent, decades has revealed appropriate results in solving major problems. Results: The results of this research indicate that the annealing approach is capable of solving such complex problems in water resources management with good precision in a reasonable period of time. Conclusions/Recommendations: The results were also compared with outputs of MODSIM which is a widely known model for solving complex water resources systems problems and benefits from "out of kilter" algorithm. The results indicate SA as a very robust and effective model in optimization of large real multi-reservoir systems.
Highlights
For a multi reservoir system, where the number of reservoirs is large, the conventional modeling by Dynamic Programming (DP) or classical Stochastic Dynamic Programming (SDP) presents difficulty, due to the curse of dimensionality inherent in the model solution
Simulated Annealing (SA) can be considered as a new method in optimization of multi-reservoir systems, especially in real large system, even with nonlinear complex objective function
Optimization and exploitation management of a multi-reservoir system is such an important issue that must be considered by planning staff
Summary
For a multi reservoir system, where the number of reservoirs is large, the conventional modeling by Dynamic Programming (DP) or classical Stochastic Dynamic Programming (SDP) presents difficulty, due to the curse of dimensionality inherent in the model solution. Simulated Annealing (SA), proposed by Kirkpatrick et al, is a randomized search method for optimization. It tries to improve a solution by walking randomly in the space of possible solutions and gradually adjusting a parameter called “temperature”. The random walk is almost unbiased and it converges to essentially the uniform distribution over the whole space of solutions. Each step of the random walk is more likely to move towards solutions with a better objective value and the distribution is more and more biased towards the optimal solutions[1]. The sequence of temperatures and lengths of time for which they are maintained is called the annealing schedule in analogy with statistical mechanics
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
