节点文献
改进的神经网络逆软测量方法在电力系统控制中的应用
Application of Improved Soft-Measurement Method Based on Neural Network Inversion in Power System Control
【作者】 李华雄;
【作者基本信息】 东南大学 , 控制理论与控制工程, 2006, 硕士
【摘要】 逆系统方法作为一种有效的非线性控制方法以其物理意义清晰、结构简单、易于实现等优点在很多领域得到了应用,但它在具体实际中通常会遇到需要反馈一些不直接可测变量的问题,这成为逆系统方法应用于具体控制问题的一大障碍。为此,本文针对逆系统控制中出现的不直接可测变量反馈问题,通过对神经网络软测量理论的改进,将其与逆系统方法相结合,给出了逆系统控制中避免不直接可测变量反馈问题的方法,并将这种方法运用到电力系统的控制中去。在国家自然科学基金的资助下,本文取得的研究成果如下:(1)对本课题组原有的神经网络软测量方法作了改进,在改进的算法中,拓展了构造“内含传感器”依据的直接可测变量的范围,使之拓展为一组过程变量的函数,并按此拓展方法给出了构造“内含传感器”子系统的具体算法,同时给出了具体算例。对原来算法所做的改进可以使该方法适用范围得以拓宽,增加了“内含传感器”构造成功的可能性。(2)为避免原有算法中易引入直接可测变量高阶导数的问题,对基于“内含传感器”逆的软测量算法步骤进行了改进。按照这种方法得到的新算法与原算法相比,构造的“内含传感器”较少引入直接可测变量的高阶导数,从而使得基于“内含传感器”逆的软测量方法在实际工程中易于实现。(3)为解决神经网络软测量方法中训练神经网络所需样本难以获取的问题,给出了一种有效的解决办法:结合逆系统控制方法,将理论上存在的函数关系式代入逆系统表达式,得到一个非线性控制律,再用神经网络逼近该复合非线性控制律,此时训练神经网络所需的样本均为直接可测变量,可以解决神经网络训练样本难以获取的困难。(4)针对多机电力系统数学模型的特殊形式,给出了多机电力系统中构造“内含传感器”逆的算法,从而使基于“内含传感器”逆的软测量方法能够运用于多机电力系统的控制中。(5)为验证本文提出的方法的有效性,按照文中提出的方法研究了电力系统单机无穷大励磁控制和多机励磁汽门控制的应用问题,通过计算机仿真验证了这些方法的有效性。
【Abstract】 As an available method, inversion method is widely applied in various field for its advantage of clear physical idea, simple structure and facile realization. However, in practical, inversion method often requires the feedback of immeasurable variable, which is an obstacle of application of inversion method. Therefore, a method for avoiding feedback of immeasurable variable is proposed in this paper by improving ANN soft-sensoring method and combining it with inversion method. Then, this method is applied in power systems control. Under the support of the National Natural Science Foundation of China, this paper obtains some progresses, as follows:(1) An improved method for ANN soft-sensoring which is presented previously by our research group. In the improved method, measurable process variables required for“assumed inherent sensor”inversion method are expanded to measurable function of process variables. Based on this method, arithmetic to construct“assumed inherent sensor”and the example of arithmetic are presented. The improvements of the method make it more possibly to construct the“assumed inherent sensor”.(2) To avoid high-order derivative of directly measurable variable, a improved arithmetic process based on“assumed inherent sensor”inversion is proposed. Compared with the previous arithmetic, the derivative order of directly measurable variable in new arithmetic is lower than that in previous arithmetic, which makes the soft-sensoring method based on“assumed inherent sensor”inversion easily to realize in practical engineering.(3) To solve the problem that the sample data for training ANN is difficult to acquired, an available method is presented.(4) An algorithm to construct the“assumed inherent sensor”inversion in multimachine is presented for the special form of mathematics model in multimachine, which makes it possible that the ANN soft-sensoring method can be applied in multimachine power systems control.(5) To verify the validity of the method proposed, in practice, we research the excitation control of power systems with one machine infinite bus and the excitation and valve control of multimachine in power systems, and the method proposed is verified by computer simulation.
【Key words】 inversion; soft-sensoring; artificial neural network; one machine infinite bus power systems; multimachine power systems;
- 【网络出版投稿人】 东南大学 【网络出版年期】2007年 04期
- 【分类号】TP183
- 【被引频次】4
- 【下载频次】279