基于递阶T-S模糊系统的软测量建模方法

Soft Sensor Modeling Based on Hierarchical T-S Fuzzy System

在采用递阶模糊系统进行软测量建模时,合理的模糊系统结构对于提高模型的性能具有重要的意义.为选择合理的系统结构,采用多目标遗传算法(MOGA)选择子系统的输入变量,并结合T-S模糊系统的特点,采用二分法划分子系统的输入空间,建立了基于递阶T-S模糊系统(HTFS)的软测量模型.该方法从结构细化、输入变量对输出的影响度、输入空间划分等多方面同时提高建模精度,简化模型结构.仿真结果表明,提出的软测量方法具有精度高、结构简单、生成规则数量少,具有良好的泛化特性等优点.

When introducing the hierarchical fuzzy system in soft sensor modeling, the rational structure of the fuzzy system is very important to the performance of the soft sensor model. To select the rational system structure, an HTFS （hierarchical T- S fuzzy system） based soft sensor model is developed by virtue of MOGA （multi-objective genetic algorithm） and dichotomy. MOGA selects the input variables of subsystem in accordance to the importance of each input variable in a sub fuzzy system, while dichotomy is used to divide the input space unevenly on the basis of the characteristics of T-S fuzzy system. The proposed method can simplify the structure of model and improve the generality, with the accuracy of the soft sensor model perfectly improved. The simulation results demonstrate the practicability and efficiency of the HTFS based soft sensor model.