摘要
该文提出一种微分进化DE(differential evolution)与误差反向传播神经网络BP(error back propaga-tion)相结合的变压器故障诊断新方法。DE算法是采用不同的策略产生变异算子,并在进化过程中采取父代和子代交叉处理的方式来提高进化速度,具有强劲的全局搜索能力,能很快寻找到全局最优点。BP神经网络具有很好的分类能力,然而其权值和阈值有收敛速度慢、易陷入局部极小值等缺点。用DE算法来优化BP神经网络的权值和阈值,可实现两种算法的取长补短。将该混合算法用于变压器故障诊断,仿真结果表明该算法具有收敛速度快、鲁棒性好、分类精度高的优点。
The proposed model combining differential evolution algorithm with BP（error back propagation） algorithm is applied to fault diagnosis of power transformer.DE algorithm uses different strategies to develop mutation operators,and during evolution it uses the approach of parent and offspring cross-processing to improve the speed of evolution.It has a strong global searching capability and can quickly find the global optimal point.BP algorithm has good ability for classification,but it has some disadvantages,such as the slow convergence of weights and thresholds learning,premature result.DE algorithm is used to optimize the weights and thresholds of BP algorithm.The hybrid algorithm is used to fault diagnosis of transformer.Results show that the proposed method has good convergence performance,good robustness and high classification accuracy.
出处
《电力系统及其自动化学报》
CSCD
北大核心
2011年第2期54-58,共5页
Proceedings of the CSU-EPSA
关键词
故障诊断
微分进化
神经网络
变压器
fault diagnosis
differential evolution
neural network
transformer
