【作者】 李之昆；

【导师】 马宏忠；

【作者基本信息】 河海大学 ， 电力系统及其自动化， 2004， 硕士

【摘要】 发电机作为电能生产的基本设备，对整个电力系统的安全稳定起着至关重要的作用。开展发电机在线监测，及时有效地诊断出发电机的故障，已逐步发展成为保证发电设备安全可靠运行的重要手段。转子绕组匝间短路是一种常见的发电机电气故障，本文对其故障机理和磁动势进行了分析，将传统的方法和智能化方法相结合，对转子绕组匝间短路的存在性和故障严重性进行了诊断。 本文的主要工作包括相互关联的四个部分： 第一部分研究发电机转子绕组匝间短路的故障机理。介绍了基于故障机理和故障特征的常用的在线诊断方法，比较了各种方法的优缺点。并详细分析了行波法的原理、实现和应用。 第二部分研究神经网络在故障诊断中的应用。通过对发电机磁动势的分析，得出发电机在故障前后运行工况不发生变化的条件下，励磁磁动势将会维持不变的结论。根据该结论构建了发电机故障诊断的神经网络，并直接获得发电机在额定运行工况下对应的故障样本，用于故障诊断，不仅对转子绕组匝间短路故障的存在性，还对故障的严重性做出了正确的判断。 第三部分研究神经网络和遗传算法相结合的方法在转子绕组匝间短路故障诊断中的应用。神经网络采用BP算法易陷入局部最优，而基本的GA易出现早熟，将BP和GA相结合用于故障诊断，并通过算例验证了该方法的可行性。 第四部分将遗传规划(GP)应用到发电机故障诊断。由于发电机的建模与参数辨识困难，发电机励磁电流和机端量之间的关系难以用精确的数学表达式来描述。利用GP算法在符号回归中的应用，得出发电机励磁电流和机端量之间的关系表达式，用于故障诊断，并通过算例验证了该方法是有效的。更多还原

【Abstract】 As the basic equipment of generating, synchronous machines play very important role in safety and stability of the whole power system. Developing on-line diagnosis to find fault in turbo-generator timely and effectively has become one of the important methods gradually to ensure the machine operating safely and reliably. Shorted rotor winding is a usual electric fault of synchronous machines. Based on the analysis of fault mechanism especially the magnetic characteristics of tum-to-turn short circuit of rotor windings in turbo generator, the paper studies the diagnose method, which combines the traditional method with the intelligent method, to identify the fault and the severity.This paper consists of four related parts as follows:The first part is concerning about the fault mechanism and the fault characteristics. The traditional on-line diagnosis methods are introduced, and the drawback of each method is put forward. At the same time, this part detailed the theory, the implementation and the application of the traveling wave method.In the second part, provided that the operating mode of the turbo generator keeps unchanged when the rotor winding appears fault, the magnetic motive force of the field winding will keep unvaried. Based on the analysis of the magnetic characteristics, an artificial nerural network can be built and the fault samples can be gained by theoretical arithmetic. The results of computation example show that the method can not only judge if shorted field turns exit but also estimate fault turns ratio.In the third part, in order to prevent shortcoming of the entrapment in local optimum of ordinary BP and the premature convergence of Basic GA, a new algorithm which combines BP with GA is presented, which is used to detect the rotor winding short. In the new algorithm, the method of coding, selection, crossover and mutation are modified. The results of computation example show that the method is feasible.In the last part, the relationship of the field current with the operational factors of the turbo generator is difficult to express exactly, belonging to the complexity of building the generator model and the difficulty of identifying generator parameters, GP algorithm is presented, which is better than other method in symbol regress. This method is shown to be effective by an example.更多还原