【作者】 周兆庆；

【导师】 陈星莺；

【作者基本信息】 河海大学 ， 电力系统及其自动化， 2005， 硕士

【摘要】 经济发展、电力市场化改革的推进对供电可靠性提出了越来越高的要求，作为电力系统故障后恢复供电的第一步，如何利用SCADA系统与保护信息管理系统提供的故障信息判断故障设备是电力调度部门迫切需要解决的问题，论文对此进行了详细的研究。 电力系统继电保护的基本工作原则是将特定设备故障对整个系统的影响限制在局部范围之内，论文以此为依据分析了输电网故障诊断的可行性，陈述了诊断的基本思路。针对故障诊断的信息需求用面向对象的方法实现了电网拓扑结线、保护配置信息的计算机表示，在通过广度优先搜索获得设备邻接信息的基础上，使用人工智能技术模拟人类专家进行故障诊断的逻辑分析过程。 论文详细研究了基于专家系统、神经网络、优化算法的三种电网故障诊断方法，从知识获取、知识存储与知识运用的角度分析了它们与故障诊断问题的结合方式，并通过算例分析了各自的优点与不足。论文综合上述三种人工智能方法的优点，提出了在故障信息与故障位置集合之间从正反向两个方向上进行推理分析的故障诊断综合法，该方法较好地权衡了计算精度与速度之间的矛盾，具有推理速度快、抗信息污染能力强的优点，不仅可以实现故障诊断的全部功能而且可以判断故障信息是否受到污染、是否具备可诊断性，具有较高实用价值。更多还原

【Abstract】 With the development of economic and the progress in electric power market the need for stable supply of power increases considerably. The automation of fault section diagnosis and restoration is also required in power systems operation so as to diminish the outage time and ensure stable supply of electric power for the customers. To identify fault sections in power systems by using information on operation of relays and circuit breakers has been an urgent problem that power enterprises must tackle. This paper is dedicated to solve this problem.The basic principle of power system relaying is to isolate the fault part from the whole, and the effective information that contributes to judging certain element’s state is implicated in its adjacent area only. This is the fundamental principle of fault diagnosis. This paper first states the feasibility and key idea of fault diagnosis, then used some appropriate methods to represent the power system in computer. Having acquired each equipment’s adjacent information, we use Artificial Intelligence methods to accomplish the diagnosis task.The frameworks of employing Expert System Artificial Neural Network and Genetic Algorithm three typical Artificial Intelligence methods to realize power network fault diagnosis are discussed in detail, the author also demonstrated these exercises in such aspects as knowledge-acquisition knowledge-storage and knowledge-exertion, examples of each methods are included. Having analyzed the characteristics of three typical AI methods, this paper proposes an integrate approach to solve the problem. The key idea of this method is to use three methods reasonably seeing to the condition of fault information. When pollution occurs we use different AI methods to infer in both directions to deal with uncertainties imposed on fault section diagnosis of power systems due to noises and loss of the information. Experimental studies for real power systems reveal usefulness of the proposed technique to diagnose fault that have uncertainty.更多还原