基于神经网络的电力计量故障诊断研究

Research on Fault Diagnosis of Power Metering Based on Neural Network

传统BP算法很难平衡干扰数据、学习的速度与效率十分迂缓、对象函数会轻易堕入局限最小值等各种问题,而高压电力计量系统对所发生故障的识别与诊断具有很高的实时性需求。针对这类问题,论文基于传统BP算法,将BP神经网络与粒子群算法相结合,并应用到高压电力计量系统的实时故障诊断中,解决了传统BP算法的各种局限性难题。首先,提取电力计量系统历史日常运行所产生的参数,筛选出故障发生时的特征参数进行归一化预处理;其次,建立一个符合需求的神经网络组织构造并设置相关参数,以此构造具有针对性的BP神经网络模型,并利用抽取的数据对该神经网络进行训练以及测试;然后,应用粒子群算法改进BP神经网络的权重数值,形成了实时性能较高的基于粒子群算法的优化神经网络电力计量故障的实时识别与诊断系统;最后,根据故障识别与诊断的测验比较结果,证明论文所设计的高压电力计量系统的故障诊断办法是能应用到实际计量系统运行当中的。

The traditional BP algorithm is difficult to balance various problems such as disturbing data,the speed and efficiency of learning is very roundabout,the object function will easily fall into the local minimum,and the high-voltage power metering system has high real-time requirements for the identification and diagnosis of faults that occur.In response to this kind of problem,the thesis is based on the traditional BP algorithm,combining the BP neural network with the particle swarm optimization algorithm,and applied to the real-time fault diagnosis of the high-voltage power metering system to solve the various limitations of the traditional BP algorithm.First of all,the parameters generated by the historical daily operation of the power metering system are extracted,and the characteristic parameters at the time of fault occurrence are screened for normalization pretreatment.Secondly,a neural network organization structure is established to meet the needs and relevant parameters are set up to construct a targeted BP neural network model,and the neural network is trained and tested with the extracted data.Then,the particle swarm algorithm is used to improve the weight value of the BP neural network to form an optimized neural network based on the particle swarm algorithm with high real-time identification and diagnosis system for power metering faults.Finally,according to the comparison results of fault identification and diagnosis tests,it is proved that the fault diagnosis method of high-voltage power metering system designed in this paper can be applied to the operation of actual metering systems.