结合贝叶斯推理与ART2wNF网络的风力发电机组偏航系统的控制策略
发布时间：2018-11-08 11:05
【摘要】：随着社会经济的发展,世界各国的能源矛盾日益突出。鉴于风能具有安全、清洁、充裕,稳定等特点,加大对风能的利用将有效地缓解能源危机和减少环境污染。而发展风力发电是利用风能的最主要的形式,其控制系统的好坏直接影响风力发电机组的效率和使用寿命,其中风电的偏航系统是实现最大化捕获风能和避免偏航机舱频繁转动的关键组成部分,这使得风力发电机组偏航系统的有效控制变得尤为重要。文章在详细分析了风力发电的偏航系统的工作原理和控制技术的基础上,提出了将结合了贝叶斯推理的ART2wNF(Adaptive resonance theory with neoteny feature)网络与风向标控制及爬山算法相结合的偏航控制策略。针对风的随机性,在风向服从正态分布的模型下,仿真得到了风向的样本数据,建立了基于最小二乘拟合下的风速模型,滤去了风信号中的噪声数据,为实现ART2wNF网络对风信号的聚类做了数据的预处理。由于ART2wNF网络在对样本进行自组织学习和聚类时,其警戒值是固定不变的,而警戒值的高低直接影响类别数的多少,为了实现ART2wNF网络警戒值的自动调节,文章应用了贝叶斯分类器的原理,在风向的正态分布模型下,计算出新的风向样本服从上一批样本分布的概率,以此后验概率作为警戒值的调节基准,设计了基于贝叶斯推理的ART2wNF网络警戒值的调整机制,提高了风信号样本的聚类效果,为解决风向在小的变化范围内出现集中风能的偏航问题奠定了基础。通过具有幼态延续特征的ART2wNF网络对经过预处理的风向数据进行自组织学习和聚类,结合风向标控制和爬山算法对ART2wNF网络的警戒值参数进行调节,得到了聚类后每个样本的聚类中心,即偏航位置,完成了自动偏航。文章通过在Matlab中搭建仿真模型模拟风力发电机组的偏航系统,仿真验证了文章提出的结合贝叶斯推理与ART2wNF网络的风力发电机组偏航系统控制策略的可行性和有效性,其不仅能有效地解决当风向在小的变化范围内(比如正负15°)出现集中风能的偏航问题,还能有效地避免偏航电机的频繁转动,对提高风能的利用率和风力发电机组的使用寿命有着重要的意义。
[Abstract]:With the development of society and economy, the energy contradiction is becoming more and more prominent. Since wind energy is safe, clean, abundant and stable, increasing the use of wind energy will effectively alleviate the energy crisis and reduce environmental pollution. The development of wind power generation is the most important form of using wind energy, and the quality of its control system directly affects the efficiency and service life of wind turbines. The yaw system of wind power is the key component to maximize the capture of wind energy and avoid the frequent rotation of yaw engine room, which makes the effective control of wind turbine yaw system become more and more important. Based on the detailed analysis of the working principle and control technology of the yaw system of wind power generation, A yaw control strategy which combines Bayesian reasoning with ART2wNF (Adaptive resonance theory with neoteny feature) network and wind vane control and mountain climbing algorithm is proposed. According to the randomness of wind, under the model of normal distribution of wind direction clothing, the wind direction sample data are obtained by simulation, and the wind speed model based on least square fitting is established to filter out the noise data in wind signal. In order to realize the clustering of wind signal in ART2wNF network, the data preprocessing is done. Because the alert value of ART2wNF network is fixed when the samples are self-organized learning and clustering, and the level of alert value directly affects the number of categories, in order to realize the automatic adjustment of ART2wNF network alert value, This paper applies the principle of Bayesian classifier, under the normal distribution model of wind direction, calculates the probability of the distribution of new wind direction samples from the last batch of samples. Based on Bayesian reasoning, the adjustment mechanism of ART2wNF network warning value is designed, which improves the clustering effect of wind signal samples, and lays a foundation for solving the yaw problem of concentrated wind energy in a small variation range of wind direction. The preprocessed wind direction data are self-organized and clustered by the ART2wNF network with the characteristics of juvenile continuation, and the alert parameters of the ART2wNF network are adjusted by combining the wind vane control and the mountain-climbing algorithm. After clustering, the clustering center of each sample, that is, yaw position, is obtained, and the automatic yawing is completed. By building a simulation model in Matlab to simulate the yaw system of wind turbine, the feasibility and effectiveness of the proposed control strategy of wind turbine yaw system based on Bayesian reasoning and ART2wNF network are verified by simulation. It can not only effectively solve the yaw problem of concentrated wind energy when the wind direction is in a small range (for example, positive or negative 15 掳), but also effectively avoid the frequent rotation of the yaw motor. It is of great significance to improve the utilization rate of wind energy and the service life of wind turbine.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM315
本文编号：2318283
[Abstract]:With the development of society and economy, the energy contradiction is becoming more and more prominent. Since wind energy is safe, clean, abundant and stable, increasing the use of wind energy will effectively alleviate the energy crisis and reduce environmental pollution. The development of wind power generation is the most important form of using wind energy, and the quality of its control system directly affects the efficiency and service life of wind turbines. The yaw system of wind power is the key component to maximize the capture of wind energy and avoid the frequent rotation of yaw engine room, which makes the effective control of wind turbine yaw system become more and more important. Based on the detailed analysis of the working principle and control technology of the yaw system of wind power generation, A yaw control strategy which combines Bayesian reasoning with ART2wNF (Adaptive resonance theory with neoteny feature) network and wind vane control and mountain climbing algorithm is proposed. According to the randomness of wind, under the model of normal distribution of wind direction clothing, the wind direction sample data are obtained by simulation, and the wind speed model based on least square fitting is established to filter out the noise data in wind signal. In order to realize the clustering of wind signal in ART2wNF network, the data preprocessing is done. Because the alert value of ART2wNF network is fixed when the samples are self-organized learning and clustering, and the level of alert value directly affects the number of categories, in order to realize the automatic adjustment of ART2wNF network alert value, This paper applies the principle of Bayesian classifier, under the normal distribution model of wind direction, calculates the probability of the distribution of new wind direction samples from the last batch of samples. Based on Bayesian reasoning, the adjustment mechanism of ART2wNF network warning value is designed, which improves the clustering effect of wind signal samples, and lays a foundation for solving the yaw problem of concentrated wind energy in a small variation range of wind direction. The preprocessed wind direction data are self-organized and clustered by the ART2wNF network with the characteristics of juvenile continuation, and the alert parameters of the ART2wNF network are adjusted by combining the wind vane control and the mountain-climbing algorithm. After clustering, the clustering center of each sample, that is, yaw position, is obtained, and the automatic yawing is completed. By building a simulation model in Matlab to simulate the yaw system of wind turbine, the feasibility and effectiveness of the proposed control strategy of wind turbine yaw system based on Bayesian reasoning and ART2wNF network are verified by simulation. It can not only effectively solve the yaw problem of concentrated wind energy when the wind direction is in a small range (for example, positive or negative 15 掳), but also effectively avoid the frequent rotation of the yaw motor. It is of great significance to improve the utilization rate of wind energy and the service life of wind turbine.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM315
【参考文献】
相关期刊论文 前2条
1 乔川;李斯阳;;1.5MW双馈风力发电机组偏航控制系统[J];控制工程;2011年S1期
2 朱亚俊;杨金明;;小型永磁风力发电系统的集成控制策略[J];通信电源技术;2010年04期
，本文编号：2318283
本文链接：https://www.wllwen.com/kejilunwen/dianlilw/2318283.html
教材专著
