改进的粒子群优化算法及其在软测量中的应用.doc

改进的粒子群优化算法及其在软测量中的应用 王改堂1，李平2,苏成利2 (1.西北工业大学自动化学院, 陕西 西安710072;　2.辽宁石油化工大学信息与控制工程学院, 辽宁 抚顺 113001) 摘要：针对粒子群优化算法不易跳出局部最优的缺点，提出了一种自适应变异的动态粒子群优化算法。该算法除了采用动态惯性权重外，还引入了自适应学习因子和新的变异算子。通过对几种典型函数的优化问题进行测试，结果表明：与动态改变惯性权重的自适应粒子群算法(DCW)算法和PSO算法相比，新算法具有很强的全局搜索能力,收敛速度和收敛精度也有所提高。然后，用该算法训练神经网络，并将其用于延迟焦化汽油干点软测量建模，仿真结果表明该算法具有较高的预测精度和良好的泛化性能。 关键词：粒子群优化算法；自适应变异；软测量；神经网络；汽油干点 中图分类号：TP202+.7 文献标识码：A An Improved Particle Swarm Optimization Algorithm and Its Application in Soft Sensor Modeling WANG Gai-tang1 LI Ping2 SU Cheng-li2 (1.School of Automation, Northwestern Polytechnical University, Xi’an Shanxi 710072, China; 2. School of Information and Control?Engineering, Liaoning ShiHua University, Fushun Liaoning 113001, China) ABSTRACT：Presented is a dynamic particle swarm optimization algorithm based on adaptive mutation, to solve the problem that particle swarm optimization algorithm is apt to trap in local minimums. Besides that it makes use of adaptive inertia weight, adaptive term coefficients and new mutation operator were introduced to the algorithm. The proposed algorithm are used to resolve the optimization problems of several standard test functions, and the result shows that the new algorithm has great global search ability, convergence accuracy and convergence velocity is also increased than PSO and DCW (adaptive particle swarm algorithm with dynamically changing inertia weigh) algorithm. Then, the algorithm is employed to train artificial neural network and applied to soft-sensing of gasoline end point in delayed coking. The simulation results show that the algorithm has better predict precision and good generalization performance. KEYWORDS：particle swarm optimization; adaptive mutation; soft sensor; neural network; gasoline end point 1引言 基金项目：辽宁省高等学校优秀人才支持计划资助(2008RC32);辽宁省高校创新团队支持计划资助(2007T103,2009T062);辽宁省教育厅科技计划资助项目(2008386)基金项目：辽宁省高等学校优秀人才支持计划资助(2008RC32);辽宁省高校创新团队支持计划资助(2007T103,2009T062);辽宁省教育厅科技计划资助项目(2008386)