A self-adaptive differential evolution algorithm design and implementation

The increasing of human population has increased the demand for almost every area of human life in the production sector, transportation sector, health sector, chemical industry, renewable energy sources, heavy industry etc. These demands have led to some problems. These problems are solved by optimization algorithms by reducing cost, increasing efficiency or making them optimum values. There are many optimization algorithm approaches to solve these problems. Most of these optimization approaches are meta heuristic algorithms. Differential Evolution (DE) Algorithm is also a meta heuristic algorithm. The DE algorithm is known for its ease of use and fast convergence. However, since the DE algorithm has very few parameters that are kept constant throughout the evolutionary process, accurately tuning these parameters is a particular problem for the performance of the algorithm. For this reason, researchers have been working on DE algorithms with self-adaptive control parameters. In this study, a self-adaptive differential evolution algorithm is designed and implemented. In the first step, a different mutation strategy was applied to the mutation phase of the differential evolution algorithm and tested on CEC 2019 benchmark set functions and compared with Whale Optimization Algorithm Modified Mutualism-WOAmM, Whale Optimization Algorithm-WOA, Moth Flame Optimization Algorithm-MFO, Butterfly Optimization Algorithm-BOA, A Sine Cosine Algorithm-SCA and JAYA algorithms in the literature. CEC 2019 benchmark functions tested on 50 dimensions. In the second step, a competitive local search strategy was added to this DE algorithm and tested on the CEC 2014 benchmark functions and compared with SHADE (Success-History based Adaptive Differential Evolution), L-SHADE (SHADE with Linear Population Size Reduction), PADE (Parameter Adaptive Differential Evolution) and LPALMDE (Parameters with Adaptive Learning Mechanism Differential Evolution) algorithms in the literature. The CEC 2014 benchmark functions was tested in 30 and 50 dimensions. According to the results, the proposed self-adaptive DE algorithm ranked first in the "average" and "best" results against the compared algorithms.