复杂多变量系统辨识算法研究及软件开发

VIP免费

3.0 侯斌 2024-11-19 4 4 2.3MB 73 页 15积分

侵权投诉

摘要

复杂多变量先进过程控制，如模型预测控制（MPC）已经在石油化工等企业成

功应用，并产生了巨大的经济效益。然而，工业实践表明，多变量过程模型辨识

实现困难而且十分费时，因此也成为了模型预测控制的一个瓶颈。这主要由于目

前MPC 工业项目的实施仍然采用开环阶跃试验来进行模型辨识，而且这些试验通

过现场工程师凭经验手动完成，这就造成了时间及人力上的高成本。模型辨识的

方法大致可分为时域辨识和频域辨识，与时域辨识相比，频域辨识不易受噪声影

响、求解过程简单，所以很多学者都致力于通过频域分析得到对象在工程范围内

允许的近似解。相比于开环辨识，闭环辨识对辨识回路对象影响小。在实际中很

多辨识对象只能在闭环下运行，不能切换到开环来进行辨识。再者，开环阶跃试

验不能充分体现多变量之间的耦合作用，而且在工业实际中，由于干扰的存在，

使得辨识结果偏差较大，从而使产品质量偏离标准。因此多变量闭环频域辨识的

方法研究显得尤为重要。

本文主要针对以上的热点问题，对复杂多变量系统辨识进行算法研究，并在此

基础上实现软件编制，从而实现针对复杂多变量系统对象的自动辨识。对于复杂

多变量系统辨识研究，本文先从单输入单输出对象入手，研究基于正常工控状态

下的闭环频域辨识方法及双模型的闭环辨识，进而对复杂控制系统串级回路的模

型辨识，并研究了针对不可操作变量的继电反馈频域辨识。在以上基础上研究了

基于正常工控下的复杂多变量系统辨识算法及针对不可操作变量的多变量系统辨

识算法。

在多变量系统辨识算法研究的基础上，本文又实现了基于算法的软件编制。软

件通过 OPC 和Modbus 通讯与底层 PLC 设备进行实时通信，获取辨识对象的输入

输出信号用于辨识。软件具有基本的信号滤波及显示功能，能保存辨识回路的基

本信息及辨识结果便于以后辨识直接加载。软件还具有对在线辨识采集的数据或

现场工程师采集的数据进行离线辨识的功能，用于模型辨识的反复校验及对比。

本文主要的创新性研究成果如下：

(1)在基于正常运行工况下的闭环频域辨识算法研究的基础上，分析双模型频率

响应，从而实现双模型的闭环频域辨识的算法拟合；

(2)分析复杂控制系统串级回路对象及继电反馈实验对象的输入信号特性，提出

针对以上两种对象的频域响应特性求解方法及模型拟合方法；

(3)研究复杂多变量系统辨识闭环频域算法，提出针对含不可操作变量回路的基

于继电反馈辨识方法，通过实验对象的信号分析，确定其频率响应特性，最终实

现模型辨识；

(4)在算法研究的基础上，实现复杂多变量对象的自动化辨识软件，包括 OPC、

Modbus 接口实现及 ADO 技术实现，还包括软件的 UI 设计、功能设计等。

关键字：多变量系统模型辨识频率响应闭环辨识辨识软件

III

ABSTRACT

Complex multivariable advanced process control, such as model predictive control

(MPC) has been successfully applied in petrochemical enterprises and has had a huge

economic efficiency. However, industrial practice indicates that the identification of

multivariable process model is the most difficult and time-consuming task of

implementing MPC project and it’s a bottleneck in the application of MPC. Because the

current implementation of MPC industrial projects still adopts open loop step tests for

model identification, and these tests were completed through on-site experience

manually, which caused the high cost of time and manpower. Methods for model

identification can be divided roughly into time-domain identification and

frequency-domain identification. Compared with the identification in time domain,

frequency-domain identification is less affected by noise and the solving process is easy,

so many researchers are devoted to the frequency-domain analysis of engineering for

the approximate solution within the scope permitted. Compared to the open-loop

identification, closed-loop identification has less influence on the object circuits. In

practice, many objects can only be identified under closed loop, and can’t be switched

to open loop. Furthermore, open-loop step test can’t fully reflect the coupling effect

among variables, and it will makes the error of identification larger due to the existence

of interference in industrial practice, so that product quality will be deviated from the

standard. Therefore, it is particularly important to research the method of multivariable

identification in frequency domain.

This article mainly aims to research the multivariable system identification about

the hot issues above, and realize the software based on those algorithms, in order to

achieve the automatic identification of complex multivariable system objects. This

paper first researched closed-loop identification methods in frequency domain under the

condition of normal control from the SISO subjects and multi-models based on

closed-loop model identification. After that, its aims are the identification of cascade

loop in the complex control system and identification by relay for inoperable variable in

frequency domain. On the basis of the above research, complex multivariable system

identification algorithm based on the normal control and identification algorithm for

inoperable variables were researched.

Based on the study of multivariable system identification algorithm, this paper also

realized the software. Software communicates with PLC by using OPC device or

Modbus communication to get input and output signal of object for identification. The

software has basic functions such as signal filtering and display of them. Some basic

information of circuits and identification results can be kept for later identification or

direct loading. The software also has off-line identification functions using data from

online identification or collected by field engineer, in order to repeatedly check and

contrast identification model.

The main innovative achievements of this paper are as follows:

(1) On the basis of researches about identification algorithms in closed loop

frequency domain under normal operating conditions, analyze frequency response of

multi-model, so as to realize the multi model of closed-loop identification in frequency

domain;

(2) Analyze the characteristics about input signal of objects in complex control

system--cascade loop and relay feedback experiment, solution methods of frequency

response and model fitting methods are presented in view of the above two kinds of

object characteristic;

(3) Research identification algorithms of complex multi variable system in closed

loop frequency domain, proposed identification methods based on the relay feedback

loop and containing inoperable variables. Determine the frequency response

characteristics through the signal analysis of experiment object to realize the model

identification finally;

(4) based on those theories, realize automatic recognition software to indentify

complex multi variable objects, including OPC, Modbus interface and ADO technology.

Software also includes UI design, software function design.

Key Word: Multivariable system, Model identification, Frequency

response, Closed-loop identification, Software identification,

中文摘要 ........................................................................................................................... I

ABSTRACT ....................................................................................................................III

第一章绪论 ............................................................................................................... 1

1.1 课题研究背景 ......................................................................................................1

1.2 论文的研究内容及方向 ......................................................................................1

1.2.2 课题来源 ....................................................................................................1

1.2.2 主要研究内容 ...........................................................................................2

1.2.3 论文体系结构 ............................................................................................3

1.3 本章小结 .............................................................................................................4

第二章闭环频域辨识算法研究 ................................................................................. 5

2.1 前言 .....................................................................................................................5

2.2 双模型频率响应 .................................................................................................6

2.3 两种模型的频域分析 .........................................................................................7

2.4 偏置继电反馈频率特性分析 .............................................................................7

2.5 传递函数模型 .....................................................................................................9

2.5 仿真实验 ...........................................................................................................10

2.6 本章小结 ...........................................................................................................12

第三章串级回路辨识算法研究 ............................................................................... 14

3.1 前言 ....................................................................................................................14

3.2 过程对象频率响应特性 ....................................................................................14

3.3 传递函数模型 ....................................................................................................15

3.4 仿真实验 ............................................................................................................16

3.5 本章小结 ...........................................................................................................17

第四章多变量闭环辨识算法研究 ........................................................................... 18

4.1 前言 ....................................................................................................................18

4.2 多变量闭环频域辨识算法 ................................................................................19

4.3 偏置继电反馈信号分析 ....................................................................................22

4.4 传递函数矩阵模型 ...........................................................................................22

4.4.1 多变量闭环辨识模型 .............................................................................22

4.4.2 基于改进继电反馈多变量辨识模型 .....................................................24

4.5 仿真实验 ............................................................................................................24

4.6 本章小结 ............................................................................................................27

第五章复杂多变量系统辨识软件开发 ................................................................... 28

5.1 设计思想 ............................................................................................................28

5.3 ModBus 和OPC 接口设计 ............................................................................... 30

5.3.1 ModBus 接口 ...........................................................................................30

5.3.2 OPC 接口 .................................................................................................35

5.4 ADO 技术实现 .................................................................................................. 37

5.4.1 ADO 简介 ................................................................................................ 37

5.4.2 使用方法 .................................................................................................37

5.5 软件设计 ............................................................................................................39

5.6 软件应用 ............................................................................................................41

5.6.1 软件基本操作 .........................................................................................41

5.6.2 软件在线实验辨识 .................................................................................48

5.6.3 离线数据辨识 .........................................................................................51

5.9 本章小结 ............................................................................................................53

第六章总结和展望 ................................................................................................... 54

6.1 工作总结 ............................................................................................................54

6.2 后续研究工作展望 ............................................................................................54

参考文献 .........................................................................................................................64

在读期间公开发表的论文和承担科研项目及取得成果 .............................................68

致谢 .............................................................................................................................69

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

I摘要复杂多变量先进过程控制，如模型预测控制（MPC）已经在石油化工等企业成功应用，并产生了巨大的经济效益。然而，工业实践表明，多变量过程模型辨识实现困难而且十分费时，因此也成为了模型预测控制的一个瓶颈。这主要由于目前MPC工业项目的实施仍然采用开环阶跃试验来进行模型辨识，而且这些试验通过现场工程师凭经验手动完成，这就造成了时间及人力上的高成本。模型辨识的方法大致可分为时域辨识和频域辨识，与时域辨识相比，频域辨识不易受噪声影响、求解过程简单，所以很多学者都致力于通过频域分析得到对象在工程范围内允许的近似解。相比于开环辨识，闭环辨识对辨识回路对象影响小。在实际中很多辨识对象只能在闭环下运行，不能...

展开>> 收起<<

复杂多变量系统辨识算法研究及软件开发.pdf

共73页,预览8页

还剩页未读，继续阅读

复杂多变量系统辨识算法研究及软件开发

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: