Abstract: To improve the optimum speed and optimization accuracy of Differential Evolution Algorithm (DEA), an improved DEA was proposed. In this algorithm, a new mutation operator following the Laplace distribution was used during the mutation, and both the mutation strategy and the crossover probability could be gradually self-adapted to fit different phases of evolution by learning from their previous successful experience. Experimental studies were carried out on five classical Benchmark functions, and the computational results show that the algorithm has faster convergence, higher accuracy and stronger robustness, and it is suitable to solve high-dimensional complex global optimization problems.

Key words: differential evolution, Laplace distribution, mutation strategy adaptation, crossover probability adaptation

摘要： 为了提高差分进化算法(DEA)的收敛速度和寻优精度，提出了一种改进的差分进化算法。在该算法中，引入了基于Laplace分布的变异算子，并且能根据以往的进化经验自适应地调整进化策略及交叉概率以适应不同阶段的进化。通过5个典型Benchmark函数的测试结果表明，该算法的收敛速度快、求解精度高、鲁棒性较强，适合求解高维复杂的全局优化问题。

关键词: 差分进化, Laplace分布, 进化策略自适应, 交叉概率自适应