基于不完全退火的正态云禁忌搜索算法及应用

Tabu search algorithm based on incomplete annealing with normal cloud model and its application

针对RC电路参数优化应用时的局限性,提出一种基于不完全退火的正态云禁忌搜索算法。该算法对原始禁忌搜索算法在寻优过程中出现的对寻优起点过度依赖、早熟、灵敏度不一致等问题进行了改进,首先通过引入改进的不完全模拟退火Metropolis准则作为算子,使得原始禁忌搜索不再依赖初始可行解,具备了一定逃离局部最优解能力;然后采用正态云模型,提出了一种自适应禁忌表记忆策略及遗忘机制,使得禁忌搜索算法具备更强随机性与模糊性,进一步增强其全局搜索性能;最后针对灵敏度不一致问题,提出串行多维多灵敏度混合编、解码策略,实现在同一算法框架下对不同寻优对象定义不同搜索范围,使得低灵敏度解受忽略程度大大降低。采用8个基准函数及TSP-Oliver30标准测试算例进行性能评估,并与其他6种同类型算法对比,证明该算法对原始禁忌搜索算法优化效果良好。将该算法应用于SiMOSFET加速器磁铁开关电源参数优化后,经仿真与实验,结果表明:所提改进算法能够对工程应用进行实际优化,具备优越性与可行性。

This study addresses the limitations in the application of RC circuit parameter optimization and introduces a Normal Cloud Tabu Search algorithm based on incomplete annealing.The algorithm improves upon the traditional Tabu Search algorithm,which excessively depends on the initial feasible solution and suffers from premature convergence and inconsistent sensitivity.The improvement begins with the integration of an enhanced incomplete simulated annealing Metropolis criterion as an operator,which enables the algorithm to escape local optima without relying on the initial solution.Next,a Normal Cloud model is employed to devise an adaptive memory strategy for the Tabu list along with a forgetting mechanism,enhancing the algorithm's randomness and ambiguity,thereby strengthening its global search capabilities.To address the inconsistency in sensitivity,a serial multi-dimensional multi-sensitivity mixed encoding and decoding strategy is proposed.It allows for different search scopes for different optimization targets within the same algorithmic framework,significantly reducing the neglect of low-sensitivity solutions.Performance assessment through eight benchmark functions and the TSP-Oliver30 standard test case,compared with six other similar algorithms,demonstrates the superior optimization effect of the proposed strategies on the original Tabu Search algorithm.When applied to the parameter optimization of Si MOSFET accelerator magnet switch power supply,simulations and experiments reveal that the improved algorithm can perform practical optimization for engineering applications,exhibiting excellence and feasibility.