多策略融合的改进黏菌算法及其应用

Improved slime mould algorithm with multistrategy integration and its application

针对黏菌算法存在自适应能力有限,抗停滞能力弱等不足,提出多策略融合的改进黏菌算法。采用Bernoulli混沌初始化,丰富种群多样性,提升算法优化精度和收敛速度;提出动态非线性递减策略,动态调节黏菌个体探索幅度,协调并优化算法全局搜索与局部开发能力;结合麻雀算法的预警机制与折射反向学习策略,优化黏菌个体分离觅食过程,防止前期优质个体流失以及后期种群多样性匮乏,提升算法整体抗停滞能力。通过对基准测试函数及部分CEC2017测试函数进行寻优对比实验,测试结果表明改进算法具有更好的寻优精度、稳定性。利用改进算法优化XGBoost参数并将其用于变压器故障诊断,进一步验证了改进策略的有效性及算法的工程实用性。

In order to address the disadvantages of the slime mold algorithm,such as limited adaptive ability and weak antistagnation capability,we propose an improved slime mold algorithm with multistrategy integration.Bernoulli chaos is used to generate the initial population,which enriches the population diversity,improves the optimization accuracy,and enhances the convergence speed of the algorithm.A dynamic nonlinear decreasing strategy is proposed to adjust the exploration range of slime mold individuals dynamically;this strategy coordinates and optimizes the global exploration and local exploitation capabilities of the algorithm.By combining the early warning mechanism of the sparrow search algorithm and refracted opposition-based learning,the foraging process of slime mold individuals is optimized,prevent-ing the loss of high-quality individuals in the early stage and ensuring population diversity in the later stage,thereby im-proving the overall antistagnation capability of the algorithm.Comparative optimization experiments are conducted on benchmark functions and some CEC2017 test functions,demonstrating that the improved algorithm exhibits superior op-timization accuracy and stability.Furthermore,the improved algorithm is used to optimize XGBoost parameters and is applied to transformer fault diagnosis,which further verifies the effectiveness of the improved strategy and the practical-ity of the algorithm in engineering applications.