摘要
为提高光伏阵列故障诊断精度,提出一种基于变分模态分解VMD(variational mode decomposition)和改进蜣螂算法IDBO(improved dung beetle optimizer)优化长短期记忆LSTM(long short-term memory)网络的光伏阵列故障诊断方法。首先,针对蜣螂算法DBO(dung beetle optimizer)收敛精度低且易陷入局部最优的问题,提出一种融合Levy飞行策略、T分布扰动策略及多种群机制的IDBO算法,通过与DBO、麻雀搜索算法、鲸鱼优化算法寻优测试对比,证明IDBO算法的优越性,再与LSTM结合搭建IDBO-LSTM故障诊断模型。其次,为充分挖掘故障特征,利用VMD提取故障数据多个层面的特征分量,作为IDBO-LSTM模型输入量。最后,实验对比结果表明,该方法的故障诊断准确率达到98.34%,优于其他5种模型,证明了所提方法的可行性及优越性。
To improve the accuracy of photovoltaic(PV)array fault diagnosis,a PV array fault diagnosis method based on variational mode decomposition(VMD)and improved dung beetle optimizer(IDBO)is proposed to optimize the long short-term memory(LSTM)network.First,in response to the low convergence accuracy of the dung beetle optimizer(DBO)and its tendency to fall into local optima,an IDBO algorithm which integrates the Levy flight strategy,T-distribution perturbation strategy and multi-swarm mechanism is put forward.Through a comparison with the DBO,sparrow search algorithm,and whale optimization algorithm in optimization testing,the superiority of the IDBO algorithm is demonstrated.Afterwards,an IDBO-LSTM fault diagnosis model is constructed in combination with LSTM.Second,to fully explore the fault features,VMD is used to extract feature components at multiple levels of fault data as input to the IDBO-LSTM model.Finally,through a comparison among experimental results,it is shown that the fault diagnosis accuracy of the proposed method reaches 98.34% and is superior to those of the other five models,proving its feasibility and superiority.
作者
李斌
高鹏
郭自强
LI Bin;GAO Peng;GUO Ziqiang(School of Electrical and Control Engineering,Liaoning Technical University,Huludao 125105,China)
出处
《电力系统及其自动化学报》
CSCD
北大核心
2024年第8期70-78,共9页
Proceedings of the CSU-EPSA
基金
国家自然科学基金资助项目(51674136、52104160)。
关键词
光伏阵列
改进蜣螂算法
变分模态分解
长短期记忆
故障诊断
photovoltaic PV)array
improved dung beetle optimizer(IDBO)
variational mode decomposition(VMD)
long short-term memory(LSTM)
fault diagnosis
