基于改进灰狼优化算法的区间二型TSK FLS方法在化工过程软测量中的应用  

Application of IGWO Optimization Algorithm-based Interval Type-2 TSK FLS Method in the Soft Measurement of Chemical Process

作　　者：曾钰翔 张栓 ZENG Yu-xiang;ZHANG Shuan(Tianhua Chemical Machinery and Automation Institute Co.,Ltd.;Henan JK Chemical Investment Holding(Group)Co.,Ltd.)

机构地区：[1]天华化工机械及自动化研究设计院有限公司 [2]河南晋开化工投资控股集团有限责任公司

出　　处：《化工自动化及仪表》2025年第1期83-93,共11页Control and Instruments in Chemical Industry

摘　　要：针对具有强非线性、复杂性的化工过程软测量建模问题,在区间二型TSK模糊系统(IT2 TSK FLS)的基础上,结合改进灰狼优化(IGWO)算法策略,提出IGWO-IT2 TSK FLS方法。与一型TSK模糊逻辑系统方法相比,IT2 TSK FLS方法可以同时建模个体内不确定性和个体间的不确定性,在现有误差反向传播(BP)算法训练的基础上,将IGWO算法用于模型前件参数和后件参数的设计,以进一步提高模型的预测性能。通过对灰狼优化算法进行改进,引入早熟收敛判断机制、非线性余弦调整策略、Levy飞行策略,提高算法的收敛速度并避免陷入局部最优。将IGWO-IT2 TSK FLS方法应用于脱丁烷塔的软测量实例建模中,在同等条件下,对一型TSK FLS方法以及BP算法、遗传算法(GA)、差分进化(DE)、粒子群优化(PSO)、生物地理学优化(BBO)、灰狼优化算法(GWO)等优化的IT2 TSK FLS方法进行比较,实验结果表明:IGWO-IT2 TSK FLS方法在性能上优于对比方法,证实了方法的有效性和应用潜力。Considering strong nonlinearity and complexity in the soft sensing modeling for chemical process,a IGWO-IT2 TSK FLS method was proposed based on interval Type-2 Takagi-Sugeno-Kang fuzzy logic system and combined with an improved grey wolf optimizer.Compared with the TSK fuzzy logic system,the IT2 TSK FLS method adeptly captured both intra-individual and inter-individual uncertainties.On the basis of the existing error back propagation(BP)algorithm training,the IGWO algorithm was further used to design model’s input and output so as to improve prediction performance of the model.Through improving the grey wolf optimization algorithm,the premature convergence judgment mechanism,nonlinear cosine adjustment strategy and Levy flight strategy were introduced to improve convergence speed of the algorithm and avoid its falling into local optimum.In addition,the IGWO-IT2 TSK FLS method was applied to the modeling of the soft sensing examples for a debutane tower.Under the same conditions,having the IGWO-IT2 TSK FLS method compared with the type-1 TSK FLS method,BP algorithm,GA,differential evolution(DE),PSO,biogeography-based optimization(BBO)and grey wolf optimization(GWO)was implemented,respectively.The experimental results show that,the IGWO-IT2 TSK FLS method outperforms them in performance,including its effectiveness and application potential.

关 键 词：IGWO-IT2 TSK FLS方法 脱丁烷塔 软测量建模 早熟收敛判断机制 非线性余弦调整策略 Levy飞行策略 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]