基于QPSO参数优化的LS-SVM智能标定

Intelligent Calibration of LS-SVM with QPSO-based Parameter Optimization

研究粒子群算法(PSO)优化最小二乘支持向量机(LS-SVM)的参数时,受到搜索空间有限的限制,容易陷入局部极值,直接影响罐容表的标定精确度的问题。针对该问题,作者采用量子粒子群算法(QPSO)选取LS-SVM的径向基核参数进行优化,建立了基于QPSO-LS-SVM的罐容表标定的软测量模型。仿真实验结果表明:该方法不用建立复杂的机理模型,只需利用QPSO-LS-SVM算法进行软标定,大幅度缩短了训练建模时间,提高了运行效率。

When we study the parameters of LS-SVM with PSO（particle swarm optimization） optimization,it is easy to fall into local extremes for the limitation of search space,thus directly affecting the calibration accuracy of tank capacity.To deal with this problem,the suthor made the optimization by selecting the radial basis kernel parameters of LS-SVM with QPSO（quantum-behaved particle swarm optimization）and built a soft-sensing model of calibrating tank capacity based on QPSO-LS-SVM.The simulation results show that it is a simple method with high precision efficiency.