The derivation of a diagonally loaded sample-matrix inversion （LSMI） algorithm on the busis of inverse matrix recursion （i.e.LSMI-IMR algorithm） is conducted by reconstructing the recursive formulation of covarian...The derivation of a diagonally loaded sample-matrix inversion （LSMI） algorithm on the busis of inverse matrix recursion （i.e.LSMI-IMR algorithm） is conducted by reconstructing the recursive formulation of covariance matrix. For the new algorithm, diagonal loading is by setting initial inverse matrix without any addition of computation. In addition, a corresponding improved recursive algorithm is presented, which is low computational complexity. This eliminates the complex multiplications of the scalar coefficient and updating matrix, resulting in significant computational savings. Simulations show that the LSMI-IMR algorithm is valid.展开更多
中点钳位(neutral point clamped,NPC)型三电平逆变器并网工作环境恶劣,IGBT面临单管与双管同时故障的挑战,这使得故障特征之间的差异变得非常微弱,进而导致双管故障的识别精度难以有效提升。为此,提出了一种新的故障诊断方法,该方法结...中点钳位(neutral point clamped,NPC)型三电平逆变器并网工作环境恶劣,IGBT面临单管与双管同时故障的挑战,这使得故障特征之间的差异变得非常微弱,进而导致双管故障的识别精度难以有效提升。为此,提出了一种新的故障诊断方法,该方法结合了多通道的二维递归融合图和轻量化多尺度残差(lightweightmultiscale convolutional residuals,LMCR)网络。首先,通过仿真获取三相电流信号作为故障信号;再利用递归图(recurrence plot,RP)将三相电流信号分别转化为二维图并进行多通道融合,以捕捉时间序列中的周期性、突变点和趋势等特征;最后,将递归融合图作为输入,输入到LMCR模型中进行故障识别,LMCR模型整合多级Inception结构和残差网络,用于提取不同尺度的特征并融合这些特征,从而保证网络的梯度消失和爆炸。实验结果显示,该方法在IGBT故障识别中表现出色,无噪声环境下平均识别准确率达100%,噪声环境中也达到了92.53%,充分证明了该方法具有较强的特征提取能力和优异的抗噪性能。展开更多
保护测量回路是电力系统继电保护的基石,其误差评估对电网安稳运维举足轻重。针对保护测量回路静态隐藏误差可能诱发保护误动/拒动的风险且难以在线监测问题,提出了一种基于递推主元分析和改进灰狼算法优化极限学习机(recursive princip...保护测量回路是电力系统继电保护的基石,其误差评估对电网安稳运维举足轻重。针对保护测量回路静态隐藏误差可能诱发保护误动/拒动的风险且难以在线监测问题,提出了一种基于递推主元分析和改进灰狼算法优化极限学习机(recursive principal component analysis and extreme learning machine optimized by grey wolf optimization,RPCA-GELM)数据驱动的保护测量回路误差评估方法。首先基于电力系统正常运行下历史数据与实时数据,应用RPCA技术在线更新主元特征模型以缩短评估时间,进一步引入4种统计算法生成4类误差监测特征量,构建误差综合评判方法进行特征优选,提升误差评估准确率。然后针对模型评估精度取决于关键参数C、σ,引入国际无限折叠混沌映射策略对灰狼算法进行优化,以提升参数寻优精度和收敛速度,在此基础上结合ELM算法提出了基于GELM的保护测量回路误差评估方法。最后通过多组对比实验验证了所提方法能实现模型性能优化,且相对其他方法有效提升了保护测量回路误差评估准确率与精度。展开更多
文摘The derivation of a diagonally loaded sample-matrix inversion （LSMI） algorithm on the busis of inverse matrix recursion （i.e.LSMI-IMR algorithm） is conducted by reconstructing the recursive formulation of covariance matrix. For the new algorithm, diagonal loading is by setting initial inverse matrix without any addition of computation. In addition, a corresponding improved recursive algorithm is presented, which is low computational complexity. This eliminates the complex multiplications of the scalar coefficient and updating matrix, resulting in significant computational savings. Simulations show that the LSMI-IMR algorithm is valid.
文摘保护测量回路是电力系统继电保护的基石,其误差评估对电网安稳运维举足轻重。针对保护测量回路静态隐藏误差可能诱发保护误动/拒动的风险且难以在线监测问题,提出了一种基于递推主元分析和改进灰狼算法优化极限学习机(recursive principal component analysis and extreme learning machine optimized by grey wolf optimization,RPCA-GELM)数据驱动的保护测量回路误差评估方法。首先基于电力系统正常运行下历史数据与实时数据,应用RPCA技术在线更新主元特征模型以缩短评估时间,进一步引入4种统计算法生成4类误差监测特征量,构建误差综合评判方法进行特征优选,提升误差评估准确率。然后针对模型评估精度取决于关键参数C、σ,引入国际无限折叠混沌映射策略对灰狼算法进行优化,以提升参数寻优精度和收敛速度,在此基础上结合ELM算法提出了基于GELM的保护测量回路误差评估方法。最后通过多组对比实验验证了所提方法能实现模型性能优化,且相对其他方法有效提升了保护测量回路误差评估准确率与精度。
