A Self-Adaptive Population-Based Hybrid Optimisation Technique for Multireservoir Benchmark Problems

Owing to the random nature of inflows and demands, including environmental flows, reservoir operation and management in water resource engineering are challenging. The complexity of this problem increases for multireservoir systems, and efficient optimisation techniques are needed to satisfy all objective functions. This study presents a control-parameter-free self-adaptive population hybrid teacher, learner phase and Rao (Rao-1, Rao-2, Rao-3, and Rao-4) (SAPHTLR) algorithm. The population size in the proposed algorithm is adjusted dynamically according to the fitness value throughout the search process. The population is divided randomly into six subgroups (two subgroups for the T and L phases and the remaining four for Rao-1 to Rao-4). Each subgroup is randomly allocated to a unique perturbation equation that directs the solution method to explore various areas within the search space. The algorithm is tested on two well-known multireservoir optimisation benchmark problems (BPs) in continuous and discrete domains. The performance of the proposed algorithm is assessed through individual comparisons with the teaching–learning-based optimization (TLBO) and Rao-1 to Rao-4 algorithms using the TOPSIS method, and the results indicate that the SAPHTLR algorithm ranks first for both BPs. The effectiveness of the proposed algorithm is also assessed against different methods in the literature. The best fitness values obtained from SAPHTLR for the discrete-time four-reservoir benchmark problem (DFRBP) and continuous-time four-reservoir benchmark problem (CFRBP) are 401.46 and 308.57, respectively. The proposed algorithm is generic and can be applied in multiobjective optimisation for real-time water resource engineering problems.