论文总字数：33130字

摘 要

近年来，为了改善全球能源短缺、环境恶化的现状，人们将目光投向可再生能源。其中，风电凭借安全环保、装机灵活等优势，逐步成为发展重点。但随着优质陆地资源开发殆尽，全球开发热点已逐步转移为海上风电，而海上运行环境特殊、投建与并网等工作复杂，需要进行准确详尽的建模分析。目前针对海上风电单机建模的研究已有一定进展，但由于风电系统动态带宽较大，原有方法难以实现对模型参数的准确辨识，且海上噪声复杂、数据量测条件有限，这都增加了参数辨识的难度。本文以双馈异步电机作为研究对象，搭建其数学模型，提出利用机组运行时的小扰动响应数据训练BP神经网络来辨识模型参数的智能化建模新方法，最后通过样本数据验证可行性。本文主要工作如下：

首先，本文简要介绍了双馈异步电机的基本结构以及运行原理，研究了双馈异步电机主要部分的数学模型，其中包括风力机模型、传动系统模型、发电机模型以及控制器模型。

其次，在Simulink中搭建含双馈异步电机的系统仿真模型，并以数值模拟出的湍流风速为小扰动激励得到风电机组的三种响应信号：转子转速、有功功率和无功功率。为了得到响应信号的特征数据，对信号进行功率谱估计并且计算响应信号典型频域特征和分析功率谱显著周期下的密度值。同时，对机组模型参数进行可辨识性分析，根据功率谱灵敏度选择灵敏度较优的参数作为辨识对象。

最后选用BP神经网络进行参数辨识，利用仿真实验得到的大量响应信号功率谱特征数据与待辨识参数构成的输入输出样本对神经网络进行学习训练，构建出能够表征二者间非线性关系的模型从而完成智能化辨识。最后，利用剩余样本数据对网络的辨识效果检验，总体平均误差在5%以内，表明该辨识方案的可行性，从而为解决实际工程中的海上风电单机建模问题提供了新思路。

关键词：双馈电机 湍流风速 功率谱估计 参数辨识 BP神经网络

Research on intelligent modeling method of offshore

wind power system

ABSTRACT

In recent years, in order to improve the current situation of global energy shortage and environmental deterioration, people have focused on renewable energy. Among them, wind power by virtue of safety and clean energy, flexible installation and other advantages, gradually become the focus of development. But with the development of high quality land resources exhausted, the global development hot spot has gradually transferred to offshore wind power. However, the offshore operation environment is special, and the construction and grid-connected work is complicated, which requires accurate and detailed modeling analysis. At present, some progress has been made in the study of single-machine modeling of offshore wind power, but because of the large dynamic bandwidth of wind power system, the original method is difficult to realize the accurate identification of model parameters, and the offshore noise is complex and the data measurement conditions are limited, which increases the difficulty of parameter identification. Based on the mathematical model of doubly-fed induction generator, a new method of intelligent modeling is proposed to identify model parameters by training BP neural network of small disturbance response data during unit operation. Finally, the feasibility is verified by sample data. The main tasks of this paper are as follows:

Firstly, this paper briefly introduces the basic structure and operation principle of doubly-fed induction generator, and studies the mathematical model of the main part of doubly-fed induction generator, including wind turbine model, transmission system model, generator model and controller model.

Secondly, the system simulation model of doubly-fed induction generator is built in the Simulink, and three kinds of response signals of wind turbine generator unit are obtained by using the turbulence wind speed simulated as small disturbance excitation: rotor speed, active power and reactive power. In order to obtain the characteristic data of the response signals, the power spectrum is estimated and the typical frequency domain characteristics of the response signal and the density value under the significant period of the analysis power spectrum are calculated. At the same time, the identification of the parameters of the unit model is analyzed, and the parameters with better sensitivity are selected as the identification object according to the sensitivity of the power spectrum.

Finally, BP neural network is selected for parameter identification. A large number of characteristic data of response signal power spectrum and parameters to be identified constitute input and output samples. These samples are proposed to study and train the neural network, and a model that can characterize the nonlinear relationship between them is constructed to complete intelligent identification. Finally, using the remaining sample data to test the identification effect of the network, the overall average error is within 5%, which indicates the feasibility of its parameter identification, thus providing a new idea for solving the problem of single-machine modeling of offshore wind power in practical engineering.

Key words: doubly-fed induction generator; turbulent wind speed; power spectrum estimation; parameter identification; BP neural network

目 录

摘 要 I

ABSTRACT II

第一章 绪论 1

1.1课题背景 1

1.1.1风能的开发现状 1

1.1.2海上风电的发展现状 1

1.1.3课题的研究方向和意义 2

1.2课题研究现状 2

1.2.1双馈异步风机建模研究现状 2

1.2.2当前存在的问题 4

1.3本文主要工作 4

第二章 双馈异步风电机组的数学模型 6

2.1风电机组的基本结构及工作原理 6

2.2空气动力学模型 7

2.3轴系模型 7

2.4双馈异步电机数学模型 8

2.4.1三相静止坐标系下的双馈电机模型 8

2.4.2两相旋转坐标系下的双馈电机模型 10

2.5控制器模型 12

2.5.1 有功功率控制 12

2.5.2无功功率控制 12

2.6本章小结 13

第三章 类噪声激励下的风电机组仿真及数据分析 14

3.1湍流风速 14

3.1.1湍流风速数学模型 14

3.1.2湍流风速数值模拟 15

3.1.3对功率谱的补充 16

3.2 双馈风电机组仿真 17

3.2.1 仿真平台的搭建 17

3.2.2仿真结果 18

3.3响应数据的功率谱估计 19

3.3.1功率谱估计方法简述 20

3.3.2小波去噪 21

3.3.3对分别进行功率谱估计 22

3.4 功率谱特征数据提取 24

3.5功率谱的显著周期分析 25

3.5.1显著周期检验方法 25

请支付后下载全文，论文总字数：33130字