基于递归最小二乘支持向量机的动态系统建模研究

Dynamic system modeling based on recurrent least squares support vector machines

针对递归最小二乘支持向量机的递归性易导致建模中偏微分方程组求解困难的问题,提出用解析法求解偏微分方程组,实现了完整的递归最小二乘支持向量机模型.首先分析了各参数的相关性,然后推导出偏微分方程的解析表达式并求解.仿真实例表明,在动态系统建模中,该模型的性能比常用的串并联模型以及现有不完整递归最小二乘支持向量机模型的精度更高、性能更好.

For the problem that recurrence leads to the difficulty in finding a solution to partial differential equations in recurrent least square support vector machines regression（RLSSVM）, analytical method is proposed to solve the equations, in which recurrence is taken into consideration, so that a complete recurrent least square support vector machines regression is achieved. Relevance between variables is examined, and the partial differential equations are solved in analytical way. Simulation results show that the recurrent model is superior to series-parallel model or other RLSSVM models proposed in recent papers, especially in modeling a system affected by noise.