Diversity enhancement-based Differential Evolution with a novel perturbation strategy

Differential Evolution (DE), an advanced population-based optimization algorithm, has been widely adopted to deal with complicated real-parameter optimization problems. However, the original DE lacked sufficient exploration capabilities and did not use population diversity mechanisms to enhance performance. To mitigate this deficiency, a diversity enhancement-based differential evolution with a novel perturbation strategy (DENPS) is proposed. Firstly, a bi-stage parameter control strategy is proposed to adjust search behavior for different stages of evolution, thus achieving a better balance between exploration and exploitation. Secondly, a perturbation strategy based on a logarithmic spiral equation is incorporated into crossover operation as a supplementary strategy to the trial vector generation scheme to maintain population diversity. Lastly, a population diversity enhancement strategy based on the covariance matrix is developed to locate and update stagnant individuals in the population. CEC2014, CEC2017, and CEC2022 test suites are used to verify the effectiveness of DE-NPS. The experimental results demonstrate that DE-NPS can obtain highly competitive performance compared to other powerful algorithms regarding optimization accuracy and convergence rate. In addition, DE-NPS is applied to a real-world optimization problem and yields satisfactory results.