基于云遗传算法优化的BP神经网络的配电网故障定位

发布时间：2018-01-13 18:32

本文关键词：基于云遗传算法优化的BP神经网络的配电网故障定位　出处：《湖北工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：当配电网中发生发生接地故障时,系统的电压会大幅的降低,给社会生产和人们的生活带来很大的影响。如何保障配电网的安全可靠,成为电力行业从业者必须要面对的课题。但是,随着用电种类的增多,配电网的电压等级不断的升高,配电网的结构也越来越复杂。虽然国家电网公司已经在加大投入成本进行电网的升级,但是由于线路的老化、人为的操作失误或者自然灾害等原因使配电网不发生故障是不可能的。因此,如何提高配电网故障诊断速率和效果一直是一个热门的研究课题。本文对配电网故障诊断的方法进行了总结,发现在配电网故障诊断中应用较多的BP神经网络存在这样的一些应用问题。BP神经网络虽然具有人工智能的优点,一般能够在进行充分的学习后能满足故障诊断的需求。但是其在诊断过程中可能出现诊断速度不够快,诊断准确性不能够达到要求,同时还可能会陷入局部最优解的情况。针对BP神经网络在配电网故障定位中存在的问题,本文中将用在全局搜索领域应用效果较好的遗传算法与BP神经网络相结合,通过改变BP神经网络在建立模型时初始权值和阈值的方法来提高其效果。同时,在本文中对遗传算法也进行了改进,在遗传算法中引入能有效处理模糊信息的云理论,形成云遗传算法,使得遗传算法本身的效果也进行了改进提高。为了验证改进后的BP神经网络算法的效果。本文首先将标准的BP神经网络应用在一简单的配电网的系统中,通过matlab仿真软件进行仿真,说明了其在配电网故障诊断中的有效性。同时再将通过云遗传算法改进后的BP神经网络应用在同一配电网系统中,通过比较两者的训练曲线图,以及最终诊断的实际输出值,可以发现改进后的BP神经网络的诊断的速率要比标准的BP神经网络的速率要快很多。然后,本文再通过针对同一BP神经网络的模型,增加其学习样本,再通过matlab仿真发现,由于学习样本数的增多,标准的BP神经网络的诊断效果要比改进后BP神经网络差很多。因此,可以得出通过云遗传算法改进BP神经网络是很有效果的,是能够提高BP神经网络在配电网故障诊断应用中的速度和精确度。
[Abstract]:When a ground fault occurs in the distribution network, the system voltage will be greatly reduced, will bring great influence to the social production and people's lives. How to ensure the safe and reliable distribution network, has become the power industry must face the topic. However, with the increase of electricity type, voltage level distribution network time the increase, the structure of the distribution network is becoming more and more complex. Although the national Power Grid Corp has increased investment in the cost of power upgrade, but due to the line of aging, causes of human errors in operation or natural disasters such as the distribution network fault does not occur is impossible. Therefore, how to improve the distribution network fault diagnosis rate and effect is always a hot research topic. This paper presents a method of fault diagnosis are summarized, found that BP neural network used in fault diagnosis of distribution network in the presence of such Although some application problems of.BP neural network has the advantages of artificial intelligence, generally can meet the needs of fault diagnosis in full after learning. But the possible diagnosis is fast enough in the process of diagnosis, the diagnostic accuracy can not meet the requirements, but also may fall into the local optimal solution. In view of the existing BP nerve network fault location in distribution networks, this paper will be used in the field of global search better application effect of genetic algorithm combined with BP neural network, to improve the effect of changing the initial weights of BP neural network in the model building and the threshold at the same time. In this paper, the genetic algorithm is improved. The introduction of cloud theory can effectively deal with the fuzzy information in the genetic algorithm, the formation of cloud genetic algorithm, the genetic algorithm itself were also improved. In order to verify the change BP neural network algorithm after effect. Firstly, BP neural network can be used in the standard system of distribution network is a simple, is simulated by MATLAB simulation software, shows its effectiveness in the fault diagnosis of distribution network. And then through the cloud genetic algorithm improved BP neural network is applied in the the same system in the distribution network, through the training curve to compare the actual output, and the final diagnosis value, can be found the rate of diagnosis of the modified BP neural network is much faster than the rate of the standard BP neural network. Then, this article through to the same BP neural network model, increase the the learning samples found by MATLAB simulation, due to the increase in the number of training samples, diagnostic efficacy of standard BP neural network is much worse than the improved BP neural network. Therefore, it can be improved by BP cloud genetic algorithm The neural network is very effective and can improve the speed and accuracy of BP neural network in the application of fault diagnosis in distribution network.

【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TM711

【相似文献】