【作者】 王红；

【导师】 荣雅君；

【作者基本信息】 燕山大学 ， 电力系统及其自动化， 2015， 硕士

【摘要】 参数辨识已成为控制科学与工程学科一门备受关注的分支,它的应用也已遍及包括电力系统、农林系统、生物系统、医药系统和经济系统,甚至社会系统等许多领域。本文主要是将参数辨识原理应用在含风电的电力系统中,主要是基于参数辨识的原理对含风电的输电线路故障测距进行了研究。主要内容如下:详细研究了电力系统参数辨识的原理及算法,并选用改进的遗传算法对参数进行辨识,将“参数估计”问题转化为“参数寻优”问题。参数辨识算法应用中的已知数据均为工频量,但电力系统故障时常会出现n次谐波,因此为了滤除谐波以及避免算法应用过程中微分方程求导用差分近似所带来的截断误差,采取了改进的傅立叶算法进行数字信号处理。参数辨识应用于风电场等值模型参数的确定上,在经容量加权单机等值方法得到双馈风电场等值模型后,选用风力发电机的三阶数学模型,并对等值后模型的可辨识性进行验证。然后在参数可辨识的情况下,以求解定子电压的误差值作为目标函数,其中的约束条件由罚函数法进行处理,将罚函数法与改进的遗传算法进行组合辨识,在一定的范围内搜索得到最优解,辨识得到双馈风电场等值模型参数。序分量有着不受系统振荡影响和负荷变化影响的优点,因此基于参数辨识的思想给出了利用瞬时序分量的故障测距辨识方法。该方法为单端测距,因此受过渡电阻和对端系统阻抗的影响。传统的输电线路对端系统阻抗可看成定值,因此可忽略其影响。但是随着风电场的并入,风电的不确定性造成对端系统阻抗对测距结果的影响越来越大。为了避免上述影响,本文利用BP(Back-Propagation)神经网络强有力的学习能力对算法得出的初始测距结果进行补偿,进而得到更为精确的解。更多还原

【Abstract】 Parameter identification has become a high-profile control science and engineering disciplines of the branch, its application has been throughout including power system, agroforestry system, biological system, medical system and economic system, and even social system and many other fields.This paper studies the parameter identification in power system applications which contains wind power, mainly related to study wind power transmission line fault location which is based on the principle of parameter identification. The main contents are as follows:Detailed study of the theory and algorithm for power system parameter identification and selection of improved genetic algorithm to identify the parameters, the "parameter estimate" is transformed into "parameter optimization" problem. Parameter identification algorithm in the amount of known data are power frequency, but the power system faults often occur n harmonics, so in order to filter out harmonic and avoid derivative algorithm in the process of differential equation with finite difference approximation of truncation error, and adopted the protection signal digital processing method.Parameter identification applied to a wind farm on the equivalent model parameters to determine, first of all use the capacity-weighted single equivalent methods doubly-fed wind farm equivalent model, followed by the choice of the third-order motor mathematical model, the model will be equivalent after identifiability analysis. Then in the case of the parameters can be identified, in order to solve the error value of the stator voltage as the objective function, the use of the penalty function method processing constraints, the penalty function method and improved genetic algorithm combined identification, within a certain range of optimal search solution, identification get doubly-fed wind farm equivalent model parameters.Based on the idea of parameter identification, the use of sequence components without affecting the system load and oscillation advantage given sequence components based on instantaneous fault location identification algorithm, The algorithm for single-end distance, so the system impedance by transition resistance, and the influence of system operation mode.Traditional peer system impedance transmission line can be seen as a given value, so you can ignore its impact. But as the wind farm is incorporated, the uncertainty of wind power ranging impact on the results of the peer system impedance is growing. In order to avoid these effects, we use BP(Back-Propagation) neural network learning ability of powerful algorithms derived from the results of the initial ranging to compensate, and then get a more accurate solution.更多还原