基于BP神经网络的低压台区单相断线故障检测方法  

BP Neural Network-Based Detection Method for Single-Phase Disconnection Fault in Low-Voltage Transformer Area

作　　者：黎镇浩 刘振东 吴宇轩 林剑雄 许兴元 刘泉辉 刘少权 陈泽铭 LI Zhenhao;LIU Zhendong;WU Yuxuan;LIN Jianxiong;XU Xingyuan;LIU Quanhui;LIU Shaoquan;CHEN Zeming(Guangzhou Power Supply Bureau,Guangzhou Guangdong 511399,China)

机构地区：[1]广州供电局,广东广州511399

出　　处：《湖北电力》2024年第5期123-131,共9页Hubei Electric Power

基　　金：广东电网有限责任公司科技项目资助(项目编号:030100KK52222032)。

摘　　要：针对低压台区单相断线故障特征量变化较小、传统判据容易失效的问题,提出了一种基于BP(BackPropagation)神经网络的单相断线故障检测方法。首先,推导了低压台区单相断线的故障特征,筛选了断线故障的正、负、零序电流突变量、三相电流、三相电压和负序电流相位变化等10个特征量,建立了传统判据的特征量集合。其次,通过主成分分析(PrincipalComponentAnalysis,PCA)方法从已构建的特征库中筛除低贡献率的特征子集,将特征集维数从10维降至3维。然后,将降维得到的特征量输入BP神经网络中进行训练,通过粒子群算法(ParticleSwarmOptimization,PSO)完成BP神经网络参数优化,增强网络训练的差异性、收敛性。最后,以某实际低压台区进行仿真算例分析,结果表明所提方法能更有效完成故障检测,并达到了100%的检测准确率,较最佳传统方法提升了8.33%的准确率。In order to solve the problem that the characteristics of single-phase disconnection fault in low voltage transformer area varies little and the traditional criteria often fail to work,this paper proposes a detection method for single-phase disconnection fault based on BP(Back Propagation)neural network.Firstly,the fault characteristics of single-phase open-circuit in low-voltage transformer area are deduced;10 characteristic quantities such as the positive,negative,and zero-sequence current mutation,three-phase current,three-phase voltage,and negative-sequence current phase change of open-circuit faults are selected,and eigen quantity set of traditional criteria is established.Secondly,by using Principal Component Analysis(PCA)method,the feature subset with low contribution rate is screened from the constructed feature database,and the dimension of the feature set is reduced from 10 to 3.Then,the feature quantity obtained by dimensionality reduction is input into BP neural network for training,and the parameter optimization of BP neural network is completed by particle swarm optimization(PSO)to enhance the differentiation and convergence of network training.Finally,a certain real low voltage transformer area is used for simulation example analysis.The results show that the proposed method can complete the fault detection more effectively,and that the detection accuracy rate reaches 100%,which is 8.33%higher than the best traditional method.

关 键 词：低压台区 单相断线 BP神经网络 粒子群算法(Particle Swarm Optimization PSO) 低压配电网 

分 类 号：TM933[电气工程—电力电子与电力传动]