改进混合蛙跳算法在蔬菜总黄酮软测量中的应用

The application of improved shuffled frog leaping algorithm in soft measurement on total flavonoids of the vegetable

针对蔬菜总黄酮化学物提取过程复杂、非线性和生物参数难以在线测量等特点,提出了基于改进混合蛙跳算法优化最小二乘支持向量机的蔬菜总黄酮软测量模型。该模型对标准混合蛙跳算法进行改进,采用反向学习的种群初始化策略,确保个体分布的均匀性;并根据群体适应度方差大小,动态调整变异概率,使算法避免陷入局部最优;最后采用改进的混合蛙跳算法对最小二乘支持向量机的参数进行寻优,实现蔬菜总黄酮软测量。仿真结果表明,基于改进混合蛙跳算法的最小二乘支持向量机软测量模型具有测量精度高,稳定性好的优点,有利于蔬菜总黄酮化学物测量工程的实际应用。

For the process of extracting flavonoid chemicals from vegetable is complex, nonlinear and its biological parameters is difficult to measure online, the soft-sensor model on total flavonoids in vegetable which adopt improved shuffled frog leaping algorithm to optimize the Least Squares Support Vector Machines was proposed. Firstly, the model which adopts reverse learning strategies in population initialization phase to ensure uniformity of the individual improved the Standard shuffled frog leaping algorithm. Then it adjusts the mutation probability dynamically according to the size of population fitness variance to make the algorithm to avoid falling into local optimum. Finally, the improved Shuffled Frog Leaping Algorithm optimizes the Parameter of least squares support vector machines to achieve soft measurement of total flavonoid in vegetable. Simulation results show that the Least squares support vector machine soft-sensor model which based on improved shuffled frog leaping algorithm has higher accuracy and better stability of measurement. It is conducive to measure the total flavonoid chemicals of vegetable in practical application.