An adaptive differential evolution algorithm based on individual-level intervention strategy for global optimization

Abstract

The differential evolution (DE) algorithm is a widely recognized metaheuristic with outstanding optimization performance and a straightforward structure. However, when DE relies exclusively on the difference information within the population to update individual positions, it can potentially cause premature convergence or stagnation, resulting in inferior performance on complex optimization problems. To enhance the optimization performance of DE effectively, we propose an adaptive DE variant, referred to as IIDE, which incorporates an individual-level intervention strategy based on a fitness state information-triggered mechanism and an opposition-based learning strategy. Furthermore, we introduce a novel mutation operator that utilizes a dynamic elite strategy and a dominant-inferior partitioning approach, along with targeted matching parameters derived from fitness state information, optimization progress information, or historical success information. To evaluate the optimization performance of IIDE, we compare it with the winner algorithm (L-SHADE) from the IEEE CEC 2014 testbed and six other high-performing DE variants developed in the past five years. The comparative results demonstrate that IIDE exhibits significant advantages in terms of statistical outcomes, optimal fitness values, and runtime efficiency.