Multi-Objective Firefly Integration with the K-Nearest Neighbor to Reduce Simulation Model Calls to Accelerate the Optimal Operation of Multi-Objective Reservoirs

Reservoirs are used as one of the surface water sources for different and often conflicting water supply purposes. Given the complex management policies governing a basin, it is essential to simultaneously consider different goals and cope with the associated trade-off in water resources management. This purpose requires coupling a multi-objective optimization algorithm with a reservoir simulation model, which this approach increases required computational efforts. Various simulation–optimization approaches have been developed and used for solving the related problems. However, they often have complicated methods and certain limitations in real-world applications. In this study, a new multi-objective firefly algorithm—K nearest neighbor (MOFA-KNN) hybrid algorithm is developed which is time-efficient and is not as complicated as previous approaches. The proposed algorithm was evaluated for both benchmark and real problems. The results of the benchmark problem showed that the execution time of the MOFA-KNN hybrid algorithm was up to 99.98% less than that of the multi-objective firefly algorithm (MOFA). In the real problem, the MOFA-KNN algorithm was linked to the 2D hydrodynamic and water quality model, CE-QUAL-W2, to test the developed framework for reservoir operation. The Aidoghmoush reservoir as a case study investigated to minimize the total released dissolved solids (TDS) and the water temperature difference between the inflow and the outflow. The results demonstrated that the MOFA-KNN algorithm significantly reduced the simulation–optimization execution time (>?660 times compared with MOFA). The minimum released TDS from the reservoir was 13.6 mg /l and the minimum temperature difference was 0.005 °C.