生物系统参数估计的改进粒子群优化算法  

Improved particle swarm optimization algorithm for parameter estimation of biological systems

作　　者：琚新刚 廉飞宇 董乐 张元 葛宏义 蒋玉英 JU Xin-gang;LIAN Fei-yu;DONG Le;ZHANG Yuan;GE Hong-yi;JIANG Yu-ying(Key Disciplines of Computer Software and Theory,College of Computer and Information Technology,Henan Finance University,Zhengzhou 450000,China;Key Laboratory of Grain Information Processing and Control of Ministry of Education,Henan University of Technology,Zhengzhou 450001,China)

机构地区：[1]河南财政金融学院计算机与信息技术学院计算机软件与理论重点学科,河南郑州450000 [2]河南工业大学粮食信息处理与控制教育部重点实验室,河南郑州450001

出　　处：《计算机工程与设计》2022年第6期1704-1710,共7页Computer Engineering and Design

基　　金：河南省高等学校重点科研基金项目(22A510014);国家自然科学基金面上基金项目(61975053)。

摘　　要：提出一种基于分解技术的粒子群优化算法,通过分解问题到每个决策变量,把PSO算法从多维索空间集中到利用单维空间上。在两个不同的模拟场景和一个由计算机模拟出的新陈代谢系统的真实数据集上,将该算法的均方根误差与迭代无迹卡尔曼滤波算法、模拟退火算法和传统PSO算法进行比较,仿真和真实数据集上的测试结果表明,该算法的均方根误差,比对比方法分别减少55.16%和19.62%,说明了提出的方法对非线性问题的未知参数估计具有较佳的性能。A particle swarm optimization(PSO)algorithm based on decomposition technology was proposed.By decomposing the problem into each decision variable,the PSO algorithm was concentrated from multi-dimensional search space to sigral-dimensional search space.The root mean squared error(RMSE)of the algorithm was compared with that of the iterative untraceless Kalman filtering,simulated annealing algorithm and traditional PSO algorithm on two different simulation scenarios and a real data set of a computer-simulated metabolic.The test results on the simulation and real data sets show that the root mean squared error of the proposed algorithm is reduced by 55.16% and 19.62%respectively,compared with that of the comparison method.It also shows that the proposed method has better performance in estimating unknown parameters of nonlinear problems.

关 键 词：生物系统建模 参数估计 迭代无迹卡尔曼滤波 模拟退火算法 粒子群优化 非线性动力学模型 时间分布 

分 类 号：TP305[自动化与计算机技术—计算机系统结构]