【作者】 王强；

【导师】 陈华；

【作者基本信息】 新疆大学 ， 控制科学与工程， 2020， 硕士

【摘要】 随着国家经济的快速发展,社会的用电量与日俱增,电力系统规模也越发庞大,电力变压器作为电网中最为关键设备之一,并且也是事故发生频率最高的设备之一,当其发生故障,会对电网的安全稳定运行造成威胁,所以及时诊断出变压器故障从而减少故障的发生,保护电网安全运行变的十分的重要。变压器在出现了故障后,会有大量气体产生并溶解在变压器油中,且不同气体含量与变压器故障类型之间存在一定的联系,利用这种关系来判断故障类型,这就是油中溶解气体分析法(DGA),针对神经网络的缺点,提出了天牛须算法(BAS)与BP神经网络算法相结合的故障诊断方法。本文先使用BP神经网络的对变压器故障进行诊断,以变压器油中溶解气体含量作为输入,以变压器故障类型作为输出,对收集到100组数据进行训练,对25组数据进行测试,发现BP神经网络存在收敛慢,容易陷入局部最优等问题,于是提出基于天牛须算法的BP神经网络故障诊断方法,通过天牛须算法来优化BP神经网络初始的权值和阈值,以优化后的权值作为初始权值,对变压器故障进行诊断。同时提出遗传算法与BP神经网络相结合的故障诊断模型。使用matlab进行仿真,发现两个模型在变压器故障诊断速度和故障诊断准确率上有所提升,弥补了BP网络收敛速度慢等缺点。在变压器故障诊断方面,提出了一种新的思路,同时也具有一定参考价值。更多还原

【Abstract】 With the rapid development of national economy,social power consumption increasing,increasingly large scale power system,power transformer,as one of the most key equipment in power grid,and is also one of the highest frequency of equipment accident,when the failure occurs,will pose a threat to the safe and stable operation of power grid,so in a timely manner to diagnose transformer faults so as to avoid or reduce the occurrence of failure is very importantTransformer after failure occurs,there will be a lot of and gases dissolved in transformer oil,and different gas content and there is a certain relationship between transformer fault type,use this relationship to judge the fault types,this is the oil dissolved gas analysis(DGA),aimed at the shortcoming of neural network,puts forward the longicorn must algorithm(BAS)combined with BP neural network algorithm of fault diagnosis methodsIn this paper,first using the BP neural network for transformer fault diagnosis,the content of dissolved gas in transformer oil as input,in the transformer fault types as output,training for the 100 groups of data collection and test for 25 sets of data,found that the BP neural network convergence rate,easily falling into the most superior,so must be based on the longicorn algorithm of BP neural network fault diagnosis methods,through the longicorn must algorithm to optimize initial weights and threshold of BP neural network,the optimized weight as initial weights,for transformer fault diagnosis A fault diagnosis model combining genetic algorithm and BP neural network is proposedMatlab is used for simulation,and it is found that the two models have some improvement in transformer fault diagnosis speed and fault diagnosis accuracy,and make up for the shortcomings of BP network such as slow convergence speed in transformer fault diagnosis,put forward a new idea,but also has a certain reference value更多还原