摘要
针对变压器故障诊断中传统BP神经网络算法准确率低、收敛速度慢、易陷入局部极小值及对初始参数较为敏感等的不足,提出一种基于蝗虫优化(GOA)算法的BP神经网络故障诊断方法。建立以变压器故障特征气体为输入、故障类别为输出的故障诊断模型,利用GOA高效的计算性能和优良的全局搜索能力对BP神经网络的权值和阈值进行参数优化。仿真结果表明,GOA优化后的BP神经网络模型相比于传统BP神经网络和基于遗传算法优化的BP神经网络,能够在保留广泛映射能力的前提下,提升网络的学习速度和全局搜索能力,进而缩短训练所需时间,提高故障诊断精度。
For the shortcomings of the traditional BP neural network algorithm in the transformer fault diagnosis such as a low accuracy,a slow convergence speed,easy to fall into local minimums and sensitivity to initial parameters,a BP neural network based on Grasshopper Optimization Algorithm(GOA)is proposed in this paper.A fault diagnosis model with transformer fault characteristic gas as input and fault category as output is established,and the weights and thresholds of the BP neural network are optimized by using the efficient calculation performance and excellent global search capabilities of GOA.The simulation results show that GOA-BP model can retain the extensive mapping capability and enhance the learning speed of the network and global search capability as compared to the traditional BP neural networks and the genetic algorithm(GA)optimized BP neural network.Thus,GOA-BP can shorten the time required for training and improve the accuracy of fault diagnosis.
作者
徐新
蒋波涛
曹雯
XU Xin;JIANG Botao;CAO Wen(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710048,Shaanxi)
出处
《电网与清洁能源》
北大核心
2021年第5期17-23,共7页
Power System and Clean Energy
基金
国家自然科学基金资助项目(11705135)
陕西省自然科学基础研究计划项目(2020JM-573)。
关键词
变压器
故障诊断
神经网络
蝗虫优化算法
transformer
fault diagnosis
BP neural network
Grasshopper Optimization Algorithm
