Hybrid algorithm of differential evolution and flower pollination for global optimization problems

Abstract

Differential evolution (DE) and flower pollination algorithm (FPA) are efficient and novel meta-heuristic algorithms, which are effective tools for solving global optimization problems (GOPs). However, a single algorithm is always restricted to its own principle and mechanism, and even suffers from the imbalance of global and local search ability when solving large-scale GOPs. The organic combination of the two algorithms with a certain strategy can improve the performance of the hybrid algorithm. Motivatedbythisidea, ahybridglobal optimizationalgorithm basedon DE and FPA is proposed. Firstly, an improved initial population generation method is providedforimproving the quality of the initial population. Secondly, an improved DE is given, and the parameters F and CR areadjustedin an adaptiveanddynamicfashion, and a combined mutation operator is designed. Thirdly, an improved FPA is proposed, and a method for adaptively and dynamically adjusting the switch probability p is given, and formulas for updating the global and local position are introduced. Lastly, a hybrid strategy of organically combining DE and FPA is proposed to form a hybrid algorithm of differential evolution and flower pollination (HADEFP). To demonstrate the performance of the hybrid algorithm, the proposed hybrid algorithm is tested against other algorithms in the literature on CEC 2017 test functions (10-D, 50-D and 100-D) and practical optimization problem, respectively. The results show that the proposed HADEFP outperforms other comparison algorithms and has better robustness.