【作者】 许凌峰；

【导师】 徐大平； 杨锡运；

【作者基本信息】 华北电力大学（北京） ， 热能工程， 2009， 博士

【摘要】 随着常规能源短缺和环境污染问题加剧,风能作为可再生绿色能源,对其开发利用十分必要。大型风力机一般都具有变桨距功能,能够通过桨距角的调节来改变机组的功率输出,使整个风场的功率控制更为主动,同时对变桨距控制技术也提出了更高的要求。受国家自然科学基金项目“大型变速风力发电混杂系统全工况优化运行控制策略研究(项目号:50677021)”的资助并作为该课题部分内容,本论文在广泛分析国内外有关资料的基础上,以变桨距型风电机组控制技术为对象,重点研究了神经网络、模糊滑模变结构等智能方法在变桨距控制中的应用,主要研究内容及创新性成果如下:1.推导了双馈发电机的能量传递关系,数学模型及其矢量控制方案。为保证双馈发电机稳定励磁,以交直交变频器为对象,分析了其数学模型和传统的网侧变换器直接电流控制策略,提出了一种新型协调控制方案。新方案中选择控制电压节点作为负载前馈补偿点,仿真表明新方案提高了母线电压调节速度。对双馈发电机励磁系统控制方案的研究和改进,使发电机系统具有稳定快速的电磁力矩响应能力,为展开变桨距功率控制技术研究提供了前提。2.分析了双馈风电机组并网过程中发电机和风力机的控制策略。双馈风电机组准同期空载并网矢量控制系统能够调节定子电压在不同转子转速下满足同期条件,但是对转子转速缺乏控制能力。为此,在分析变桨距风力机启动特性和并网升速控制要求的基础上,创新的将一种神经网络辨识与模型预测控制相结合的控制方案应用于变桨距并网转速控制。仿真结果表明新方案下转速控制效果和克服扰动能力优于传统PI控制器。3.针对变桨距风力发电系统非线性、强扰动特点及变桨距控制要求,创新的将一种基于神经网络的前馈-反馈复合变桨距控制方案应用在高于额定风速的恒功率控制中。利用神经网络进行风力发电系统建模并用作闭环控制系统前馈,以降低风速变化对系统的影响。仿真结果表明新方案有效。4.从提高控制系统鲁棒性出发,针对变桨距风力机组桨距角控制,利用反馈线性化方法,结合滑模控制和模糊控制,创新的提出了一种模糊滑模控制器,具有不需要知道被控对象精确数学模型的优点。该控制器利用模糊逻辑逼近被控对象模型,实现滑模控制律并利用李亚普诺夫函数证明了控制器的稳定性。仿真结果证明新方案具有良好控制性能。更多还原

【Abstract】 Due to continually diminishing reserves of normal energy sources and environment pollution,it is very necessary to develop wind power as its renewable and clean characteristic.Large scale wind turbines always have variable pitch,which can control output power by pitch regulating,and thus higher pitch control technology is required. Supported by the National Science Foundation of China Project ’Research on optimizing control strategy overall operating condition for variable wind turbine hybrid system(NO: 50677021)’ and as a part of it,this thesis studies intelligent control technology for variable pitch wind turbine,including neural network,fuzzy sliding mode control and their applications on pitch control,primary contents and original contributions of this thesis can be summarized as follows:Firstly,power transfer relation,mathematic model and its vector control strategy of doubly fed induction generator(DFIG) is deducted.Armed at stable operating of DFIG, mathematic model and normal control strategy for AC-DC-AC converter is studied and a novel corresponding control strategy is proposed,in which control voltage node is selected as load fed-forward position.Simulation indicates that DC-bus voltage regulating speed is increased by novel strategy.Research and improvements on DFIG exciting control strategy make generator system possess fast and stabile electric torque response ability,which supply precondition for variable pitch control technology studying.Secondly,DFIG and wind turbine control strategies for cutting-in are analyzed. DIFG cutting-in vector control system can make generator stator voltage meet cutting-in requirement with varied rotor speed,but can’t maintained rotor speed.Therefore,based on startup characteristic and cutting-in speed control requirement of wind turbine studying,a novel pitch control strategy for cutting-in speed,with neural network identification and model predictive control,is originally proposed.Simulation shows that the nonlinear and disturbance problems are solved by novel strategy,whose control performance and anti-disturbance ability is superior to normal PI controller.Thirdly,according to variable pitch wind characteristic and control requirements,a novel pitch control strategy based on neural network model fed-forward,is originally proposed for power control during high wind velocity.Wind turbine model is established by neural network and used as fed forward for closed loop control system to deduce influence of wind varying.Simulation indicates its efficiency.Fourthly,form the view of enhancing control system robustness,a novel fuzzy sliding model control strategy is proposed for pitch control,with advantage of plant model is unnecessary.Adopting feedback linearization,a fuzzy sliding mode controller is constructed combining fuzzy control with sliding mode control.By using fuzzy logic to estimate uncertainty model in control law,slide mode control law is deduced to perform effective control,and stability is proved by Lyapunov function.Simulation proves its good control performance.更多还原