基于SVR学习的多源空间分布系统3-DFLC  

SVR Learning Based 3-D Fuzzy Logic Controller Design for Spatial Distributed Systems with Multiple Control Sources

作　　者：张宪霞[1] 秦磊[1] 李佳佳[1] 

机构地区：[1]上海大学大学机电工程与自动化学院上海市电站自动化重点实验室,上海200072

出　　处：《系统仿真学报》2014年第10期2418-2422,2429,共6页Journal of System Simulation

基　　金：国家自然科学基金(61273182)

摘　　要：针对多控制源空间分布系统,提出了一种基于多输出支持向量回归机(SVR)学习的3-D模糊控制(3-D FLC)设计方法。基于多控制源空间分布系统的空间局部影响特性,将多控制源系统分解成多个单控制源子系统,每个子系统均采用单独的3-D FLC控制。将3-D FLC的空间模糊基函数用作SVR的核函数,建立起3-D FLC与SVR的数学等价关系。用SVR学习隐含有控制规律的输入输出数据集,将学习后的支持向量作为关键数据点用来构建3-D FLC的模糊规则库,完成3-D FLC的设计。将该方法应用于填充床催化反应器温度控制,仿真结果验证了该方法的有效性。A design method of 3-D fuzzy logic control（3-D FLC） was proposed for spatially distributed systems with multiple control sources based on Support Vector Regression（SVR） learning. Based on the local spatial influence feature, the system with multiple control sources was decomposed into multiple subsystems with one control sources, which would be controlled by a 3-D FLC. The spatial fuzzy basis function of 3-D FLC was taken as the kernel function of a SVR, and then an equivalent mathematical relationship was established. The SVR was used to learn the support vector from input-output data set hidden with control law, which would be taken as key data points to construct fuzzy rule base of 3-D FLC. The proposed method was applied to a catalytic packed-bed reactor and simulation results have verified its effectiveness.

关 键 词：SVR 3-D FLC 空间分布系统 空间模糊基函数 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置]