基于Hammerstein-Wiener的燃气轮机非线性建模方法研究  

Research on Nonlinear Modeling Method of Gas Turbine based on Hammerstein-Wiener

作　　者：栾俊奇 李淑英[1] 曹云鹏[1] 古成中[2] LUAN Junqi;LI Shuying;CAO Yunpeng;GU Chengzhong(College of Power and Energy Engineering,Harbin Engineering University,Harbin,China,Post Code:150001;Unit 91663 of People′s Liberation Army of China,Qingdao,China,Post Code:266012)

机构地区：[1]哈尔滨工程大学动力与能源工程学院,黑龙江哈尔滨150001 [2]中国人民解放军91663部队,山东青岛266012

出　　处：《热能动力工程》2024年第11期186-194,共9页Journal of Engineering for Thermal Energy and Power

基　　金：国家科技重大专项(J2019-I-0003-0004)。

摘　　要：对燃气轮机建模是实施控制、性能优化及状态监测等工作的基础。本文提出一种基于Hammerstein-Wiener的燃气轮机非线性建模方法,首先采用传递函数建立燃气轮机动态线性模型,然后分别采用分段线性函数和小波变换网络作为非线性输入环节和输出环节的函数,用来描述燃气轮机的非线性特性,并构建了损失函数。将模型的信息向量与参数向量解耦,基于燃气轮机实际运行数据,利用高斯牛顿非线性最小二乘法(Gauss-Newton)实现对参数向量的辨识,生成燃气轮机的非线性模型,并与ARX、BP和CNN模型进行对比,利用测试数据对非线性模型的外延性进行验证。结果表明:所建立的非线性模型具有更高的精度,均方根误差(RMSE)最大为0.0109,平均绝对百分比误差(MAPE)最大为3.3054%,拟合度(FIT)最小为0.9987,实现了对实际燃气轮机监测数据的准确拟合。The modeling of a gas turbine serves as the foundation for its control,performance optimization and state monitoring.This paper proposed a nonlinear modeling approach for gas turbines based on Hammerstein-Wiener methodology.Firstly,the dynamic linear model of the gas turbine was established using the transfer function.Then,the piecewise linear function and the wavelet transform network were employed as the functions of the nonlinear input and output links,respectively,to depict the gas turbine′s nonlinear characteristics.And the loss function was constructed.The information vector of the model was decoupled from the parameter vector,and the parameter vector was estimated using the Gauss-Newton nonlinear least squares method based on actual operational data of the gas turbine,thereby obtaining a nonlinear model for the gas turbine,which was compared with ARX,BP and CNN models.The test data was utilized to validate the nonlinear model′s extensibility.The results demonstrate that the established nonlinear model exhibits enhanced accuracy.The maximum root-mean-square error(RMSE)is 0.0109,the maximum mean absolute percentage error(MAPE)is 3.3054%,and the minimum FIT is 0.9987,indicating the model′s capability to accurately fit actual gas turbine monitoring data.

关 键 词：燃气轮机 非线性建模 HAMMERSTEIN-WIENER模型 Gauss-Newton法 

分 类 号：TK221[动力工程及工程热物理—动力机械及工程]