充分搜索多策略花授粉算法在PID参数优化中的应用

Application of Multi-strategy Flower Pollination Algorithm under Exhaustive Search in PID Parameter Optimization

PID控制器广泛应用于工业领域,结构简单,控制效果良好,参数对控制器起到了绝对作用。花授粉算法是一种应用广泛的元启发算法,但存在易陷入局部最优、迭代后期收敛速度慢、寻优精度差等不足。用加入随机扰动的反双曲正切函数充分搜索策略的转换概率替换原本的固定概率,平衡全局搜索和局部搜索;在异花授粉中引入新型动态因子,改变母系花粉位置的影响;在自花授粉阶段提出权重学习策略,让花粉向优秀花粉方向聚拢;引入翻筋斗探索策略,加大种群多样性。在此基础上提出充分搜索下多策略花授粉算法(MSFPA),对比分析了花授粉算法(FPA)、粒子群算法(PSO)、差分进化算法(DE)和象群算法(EHO)对9个测试函数的仿真实验,结果表明MSFPA算法收敛速度快,性能更优。将MSFPA算法应用于PID参数优化中,经过优化的系统超调量较小,且调整时间较短,有较强的稳定性。

PID controller is widely used in the industrial field,with simple structure,excellent control effect,and parameters play an absolute role in the controller.Flower pollination algorithm is a widely used meta-heuristic algorithm,but it has some shortcomings,such as easy to fall into local optimum,slow convergence speed in the late iteration stage,and poor optimization accuracy.The original fixed probability is replaced by the conversion probability of the full search strategy of the antihyperbolic tangent function with random perturbations,and the global search and local search are balanced.The introduction of new dynamic factors in cross-pollination,the influence of maternal pollen position is changed.In the self-pollination stage,a weighted learning strategy is proposed to make the pollen gather in the direction of excellent pollen.Somersault exploration strategies are introduced to increase population diversity.On this basis,the multi-strategy flower pollination algorithm under exhaustive search(MSFPA)is proposed,and the simulation experiments of four algorithms,namely Flower Pollination Algorithm(FPA),Particle Swarm Optimization(PSO),Differential Evolution Algorithm(DE)and Elephant Swarm Algorithm(EHO),are compared and analyzed.It is showed that the MSFPA has fast convergence speed and better performance.The MSFPA algorithm is applied to the PID parameter optimization,and the optimized system has a small overshoot,a short adjustment time,and strong stability.