A Comparison Of Different Optimization Techniques Based On A Real Water Resources Allocation Problem 'Source Title' Proceedings of Regional Conference on Water Resources Management, Isfahan, Iran, 1995
Bibliographic details
Summary
A real niuldpurpose-multireservoir system, 'Complementary Operation of Karun I, Dez, and Karun III Storages in Khozestan/Iran', is employed as a basis of comparing the three state-of-the-art network solvers, OKA, NETFLO, and RELAXT-III. First, the problem is solved using different years of data by ARSP, a simulation model based on network flow programming, which contains OKA as its network solver. Then, NEMO and RELAXT-III are modified to enable use of the same data set as OKA does and the problem is solved by the modified codes. A comparison of these three minimum cost flow techniques, based on 108 solution sets of different time periods (i.e. 12 months) and amounts of water supply (i.e., wet, medium, and dry years), indicates that the modified NETFLO is faster than OKA (up to 4.5 Limes), and the modified RELAXT-III is the fastest one (i.e. up to 10 times faster than OKA) for solving this problem. The total costs given by the different codes are identical for each time period of the different years. Also, sensitivity analysis with the amovnt of water and the cost structure (i.e. alternating the operation policy) indicates that while in wet supply period, the different codes allocate identical. flows to each arc of channels in any operation policy, in dry supply period, identity of flows in each arc allocated by different network solvers is sensitive to the cost structure (operation policy).