【作者】 张新房；

【导师】 徐大平；

【作者基本信息】 华北电力大学（北京） ， 热能动力工程， 2005， 博士

【摘要】 能源、环境是当今人类生存和发展所要解决的紧迫问题。风力发电清洁无污染,施工周期短,投资灵活,占地少,具有较好的经济效益和社会效益。由于在目前技术条件下风电与火电、水电相比从造价、电能质量、设备制造和控制技术等领域存在劣势,使得我国风电领域的理论和应用研究与发达国家存在很大差距。国内对风电技术的研究十分薄弱,风力机的大型化、变桨距控制、变速恒频等先进风电技术还远未解决,致使我国大型风力机几乎全部为进口产品。因此,深入研究风力发电的各项技术对于持久开发风能和实现风力机国产化具有重要意义。 风力发电技术是涉及空气动力学、自动控制、机械传动、电机学等多学科的综合性高技术系统工程。目前,风电领域的研究难点和热点集中在风机大型化、先进控制策略和优化技术等方面。由于风能具有能量密度低、随机性和不稳定性等特点,风力发电机组是复杂多变量非线性不确定系统,因此,控制技术是机组安全高效运行的关键。本文针对风电机组控制的相关问题展开研究,主要内容归纳如下: (1)采用分析建模和实验数据验证相结合的方法建立大型风力发电机组的非线性数学模型,以描述整台风力机的动态行为。此模型对不同控制概念的风轮具有通用性,不但能描述机组的基本动力学特性,还适合于控制目的。模型的有效性通过现场测得的风力机数据验证,并且分析了模型失配的主要原因。此模型可以用于检验控制策略的有效性。 (2)针对风速仪测量风速的不准确性,本文将大型变速变桨距风力机的有效风速估计看作一个标准的软测量问题,提出基于支持向量机的有效风速软测量。软测量技术的核心问题是建立软测量的数学模型,以实现辅助变量对主导变量的最优估计。文中推导了回归型支持向量机和最小二乘支持向量机的算法,给出基于支持向量机软测量建模的具体步骤。 (3)首次提出基于支持向量机的非线性预测函数控制算法和双馈发电机的预测函数控制。利用基于线性核函数的支持向量机进行非线性系统辨识,建立预测模型。通过预测函数控制的机理推导出采用一个基函数(阶跃函数)和两个基函数(阶跃函数和斜坡函数)两种情况下的控制律解析表达式。该算法具有在线计算量少,跟踪性能好,抗干扰能力强的特点。针对双馈发电机快速响应对象的特点,综合考虑发电机对控制系统在设定值跟踪能力、抗扰动能力和鲁棒性等方面的设计要求,基于定子磁场定向矢量控制系统模型,结合动态线性化和反馈稳态解耦技术,提出了双馈发电机有功、无功功率的预测函数控制。 (4)在分析风力机能量流动基础上,本文提出利用模糊逻辑系统得到低风速时风力机的参考转速,实现最大风能捕获。该方法不需要测量风速,避免了风速测量华北电力大学博士论文 ii的不精确性,不需要了解风力机的气动特性。(5)根据风力发电机组的运动方程,提出了风力机转速自适应最优模糊控制。算法综合考虑机组的机械特性和电气特性,系统辨识作为控制算法的一部分自动执行。在介绍自适应最优模糊控制原理的基础上,提出了一种自适应模糊逻辑系统的改进最近邻聚类学习算法,该算法在确定聚类时同时考虑输入输出信息的影响,并根据聚类样本数目的多少自适应调整衰减因子。改进算法克服了原算法中敏感参数多,不易调整的缺点。(6)提出支持向量机变桨距智能控制算法。功率系数是桨距角和叶尖速比的非线性函数,本文提出基于支持向量机的功率系数智能模型,该模型具有很好的功率系数拟合特性和较强的泛化能力,该方法对不同制造厂商的风轮具有较好的适应性和通用性。在功率系数智能模型基础上,提出变速变桨距风力机的智能控制方案。该方案包括两个协调工作的控制回路,低于额定风速时,采用自适应最优模糊控制调节发电机电磁转矩设定值,跟踪最优参考转速,实现最大风能捕获;高于额定风速时,采用支持向量机变桨距控制算法,控制机组的额定输出功率。仿真结果表明,风轮可以在变化的风速中获取最大能量并能有效改善控制器切换时引起的功率暂态响应,具有较好的实时性和鲁棒性。支持向量机首次引入风电控制领域,体现了很好的性能。关键词:风力发电机组,双馈发电机,变速恒频,变桨距控制,支持向量机,自适应模糊控制,预测函数控制更多还原

【Abstract】 Energy and environment are pressing problems that must be settled by human beings for future survival and development. Wind power has preferable economic and social benefits because of its cleanliness and free pollution, short construction period, flexible investment and few occupation of land. At present, there are big gaps between china and developed countries in wind turbine’s theory and application research, because wind power has disadvantage with thermal power and hydropower in cost, power quality, equipment manufacture and control technology. Research of wind turbine in china is very weak and many advanced wind turbine technologies are not solved such as high power capacity, pitch control and variable speed constant frequency etc. Thus, almost all advanced large-scale wind turbines are imported from overseas. In all, indepth research of wind turbine technology has very important meaning for persistent wind turbine development and home production. Wind turbine is a comprehensive system engineering of high technology which deals with aerodynamics, automatic control, mechanical drive and generator etc. The research difficulties and hot pots of wind power are focused on high power capacity, advanced control and optimization etc. Control is crucial to the efficiency and reliability of wind turbines. Wind energy has lower density, instability and randomicity, wind turbines have strong nonlinear multivariable with many uncertain factors and disturbances. The thesis expands its investigation on relevant control problems of wind turbines, main contents can be concluded the following: (1) Nonlinear mathematical model of large-scale wind turbine is established by combining analysis modeling with experimental data verification to reveal the dynamic behavior of complete wind turbine. The model has generality to different wind turbine control concepts. It reflects the basic dynamics and is suitable for control purpose. Model validity is tested by measuring field data of wind turbine. Main reasons of model mismatch are analyzed. The model can be used for the validity of control strategy. (2) The estimation of effective wind speed of large-scale variable speed variable pitch wind turbine is regarded as a standard soft sensor problem to aim at measuring inaccuracy of an anemometer. Soft sensor of effective wind speed is proposed based on support vector machine (SVM). The core of soft sensor is mathematical model to realize optimal estimation of dominant variable by assistant variables. The algorithms of SVM and least square SVM for regression are derived in the paper. The detailed steps of SVM soft sensor model are presented. (3) Nonlinear predictive functional control (NPFC) based on SVM and PFC for doubly-fed induction generator are proposed for the first time. A predictive model is established by nonlinear system identification with SVM based on linear kernel function. An explicit control law is obtained through the predictive functional control mechanism with both one base function (step function) and two base functions (step and ramp functions). The algorithm for nonlinear system has light online computational burden, good reference tracking and efficient disturbance rejection. Doubly-fed induction generator has fast dynamics. Control system design for generator must be comprehensively considered on the requirements of reference tracking, sensibility to perturbations and robustness. PFC is proposed for active power and reactive power of doubly-fed induction generator based on the stator flux-oriented vector control model combining with the dynamic linearization and decoupling state feedback. (4) On the basis of power flow analysis of wind turbine, a new method for estimating the optimal rotating speed at low wind speed is proposed by an application of fuzzy inference system. The method can realize maximum wind energy capture and avoids the inaccuracy by measuring the wind speed. It need not know the aerodynamic characteristics of wind turbine. (5) Adaptive optimal fuzzy system for speed control is proposed base更多还原