【作者】 袁小芳；

【导师】 王耀南；

【作者基本信息】 湖南大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 随着电力工业的不断发展,对于电能的需求越来越大,对供电可靠性与电能质量的要求也越来越高。发电机组是电力系统中提供电能的关键设备,对于电力系统的安全经济稳定运行有着至关重要的作用。大容量发电机组的增多并向超临界化方向发展,使得发电机控制越来越复杂、也越来越重要。论文首先介绍了电力系统控制和稳定性的一般描述以及发电机控制的重要性,并说明了发电机系统是一个复杂的非线性对象,控制难度较大。接着,从国内外发电机系统控制的研究现状出发,提炼出其中有待进一步解决的关键问题并进行深入的研究,从而形成比较系统的发电机系统智能控制理论和方法,为发电机组的高效节能与经济运行提供理论基础。针对复杂发电机系统控制中的一些科学问题,本论文结合国家"十五"重大技术装备国产化创新研制项目—“交流励磁水轮发电机及其控制系统研制”,重点开展了以下几个方面的研究工作:1、提出了基于混沌模式搜索法的同步发电机参数辨识方法,将模式搜索法与混沌优化算法相融合,给出了一种新型的混沌模式搜索法,提高了混沌优化算法的搜索效率和搜索精度。将同步发电机参数辨识问题看作参数矢量的组合优化问题,建立对应的目标函数,以优化算法来搜索最佳参数值。该参数辨识方法具有不受对象非线性特征影响、对扰动信号限制少、搜索速度快等优点。2、提出了励磁系统的一种逼近模型控制方法。首先介绍了励磁系统的一般原理和数学模型,随后提出了一种逼近模型控制器,采用基于Taylor扩展的线性化方法来得到近似的逼近模型控制律,利用径向基神经网络建模来具体实现控制律。论文分析了该逼近模型控制律的鲁棒性特征,并给出了其仿真实验结果。3、提出了一种励磁系统的逼近内模控制器。该逼近内模控制器包括逼近模型控制器和反馈补偿二个部分,采用基于Taylor扩展的线性化方法得到逼近模型控制器,并由递归神经网络建模来具体实现,再选取一个鲁棒性滤波器作为反馈环节,并分析了逼近内模控制器的鲁棒性特征,仿真实验也验证了其有效控制性能。4、水轮机调节系统是一个水、机、电的综合调节系统,针对该调节系统的非线性模型,研究了一种刚性水锤效应时的水轮机调速器的逆模型控制方法。首先介绍了水轮机的调速系统的原理,分析了其可逆性,采用支持向量机来建立逆模型,并对比研究了二种不同的逆模型控制器:直接逆模型控制和带PI控制补偿的逆模型控制。直接逆模型控制易于实现,然而鲁棒性和抗干扰性性能明显不如带PI控制补偿的逆模型控制。5、针对汽轮发电机调速系统,设计了一种汽门系统的自适应逆模型控制。首先介绍了汽轮机调速系统的调速原理和数学模型,分析了该系统的可逆性,采用最小二乘支持向量机来辨识汽门系统的模型和逆模型,在此基础上,提出了一种基于自适应学习率的逆模型控制系统,分析了该学习算法的收敛性特征。6、研究了汽轮发电机系统的综合控制,将发电机组的励磁和汽门系统相结合,设计了一种多模型控制系统。该多模型控制系统以发电机组的典型工况为依据,设计了不同工况时的子模型控制器,并提出了一种自学习算法在线调整优化控制器的控制规则。同时,建立子模型库,依据运行工况与子模型的匹配程度来决定子模型控制器的加权系数。7、介绍了用于多机组发电机系统综合控制的等效模型,并建立了三机组发电机系统综合控制的状态空间模型;随后,采用线性二次最优控制方法来设计综合优化控制器。本论文将权矩阵的选取看作一个多变量优化问题,提出一种并行混沌优化融合单纯形的优化算法,采用该优化算法来搜索最佳的权矩阵,从而为线性二次最优控制提供了一种有效方法。仿真实验也验证了该优化控制器的有效性能。论文最后总结了全文的主要创新性研究成果,对下一步研究工作进行了展望。更多还原

【Abstract】 With the development of power system industrial, the requirement of power energy is more and more, at the same time, the quality requirement of energy supplying and power stablity is higher and higher. Generators are the key equipments for energy supplying, and their performances are very crucial in power system operation. It is a tendency that the generators have higher capacity, this also means that the effective control of generators is more difficult and more important.Firstly, this dissertation introduces the outline of power system control and stability, then presents the importance of generators system control. The control of generators is complex and defficult since they are complex nonlinear plants. Subsequently, this dissertation shows the recent research development in generators control, and selcets several key problems of generators as research objectives to constitute systemic intelligent control theory and methodes. Focusing on these scientific problems, this dissertation is partially supported with the“10th Five Year Plan”Key Technology and Equipment Project-“AC excited hydrogenerator and control system development”, and has carried out the following research works:1. A novel parameters identification method for synchronous generators based on Chaotic pattern search algorithms (CPSA) is proposed. The CPSA is a combination of Chaotic optimization algorithms and pattern search algorithms, and it has good search ability and its searching is fast. Parameters identification for generators is treated as a combined optimization of parameters vectors, and an objective function is constructed for this optimization problem. The CPSA method can search the optimal parameters vectors fast, while it is not affected by the nonlinear character and restrictions for disturbance signals are less.2. An approximate model control method for excitation control is presented. Firstly, the outline and mathematic model for excitation system are shown, then an approximate model controller is presented using a direct linearization approach via Taylor expansion, and the control law is realized using RBF neural networks modeling. The robustness of the proposed excitation controller is given.3. An approximate internal model control method for excitation control is studied. Approximate internal model control inclueds approximate model control and feedback compsentation. The approximate model contrller is presented using a direct linearization approach via Taylor expansion, while the feedback compsentation is realized using a rubustness filter. The robustness of the proposed controller is given.4. Turbine governor system is a water, mechanical, electricity integrated plant. An inverse model control method for a rigid water hammer turbine governor is studied for the regulation of non-linear model. First, the principle of the turbine speed regulation system is introduced, and its reversibility is also shown. Secondly, support vector machines is used to identify its inverse model, and two kinds of different inverse model controller are compared: direct inverse model control, inverse model control with PI control Compensation. Direct inverse model control, its design is simple, yet its robustness and interference performance are not as good as inverse model control with PI control compensation.5. A new adaptive inverse model controller for the steam valve is shown in this dissertation. Firstly, the principle and the mathematical model of steam turbine governor is introduced. Secodnly, the irreversibility of the system is shown. Then least-squares supprot vector machines is used to identify the model and the inverse model, and an adaptive learning rate is used for this inverse model control systems. The convergence characteristics of the learning algorithm is analysed.6. A multiple models control system (MMCS) is proposed for excitation and turbine control of synchronous generator. Firstly sub-model control rules for synchronous generator at variable operation points are derived from different operation samples. Then a self-learning algorithms is presented for the online learning of sub-model control rules. In the same time, the sub-model is constructed and the weighting factors for sub-controller are decided by the matching degree between the operation point and the sub-model.7. The integrated control of multi-unit generating system is also presented, and three generating units integrated control system with state-space model is established. The linear quadratic optimal control method is used to design the integrated controller. This dissertation regareds the selection of the weighting matrix of linear quadratic optimal controller as a multi-variable optimization problem, and proposes a parallel chaos optimization algorithms integration with the simplex method to search the best weighting matrix. It is an effective way to reach linear quadratic optimal controller. The simulation results also show the performance of the proposed controller.In the end, the main innovations of the dissertation are summarized, and the fields for further investigation are expected.更多还原