摘要
由电力开关管组成的逆变器非线性较强,进行故障诊断比较困难。采用SAPSOBP综合诊断网络(模拟退火算法、粒子群算法和BP神经网络的有效结合)和小波分解相结合的方式,以输出侧电压作为特征信号,经小波分解后的离散近似信号和离散细节信号作为特征向量,通过模拟退火算法对粒子群算法权重和加速因子的优化,结合被粒子群算法优化了阈权值的BP网络及其分类预测功能,对特征向量诊断。由实验结果表明,SAPSO-BP网络和小波分解相结合对故障元件的诊断有良好的效果。
The inverter which is constituted by power electronic circuits has strong nonlinearity. Usually it is difficult to diagnose the fault. In this paper,the integrated Network of SAPSO-BP（ that is an effective combination of simulated annealing algorithm,particle swarm optimization algorithm and BP neural Network）is combined with wavelet decomposition,and then voltage of output is used as characteristic signal and discrete approximation signals,and discrete detail signals after wavelet decomposition are used as feature vectors. And simulated annealing algorithm is applied to optimize the weight and acceleration factors of particle swarm optimization algorithm. Combining BP Network after being optimized by particle swarm algorithm and its predicative function,to diagnose feature vector. The experimental results show that the combination of SAPSO-BP Network and wavelet decomposition has a good effect on fault diagnosis.
作者
李从飞
田丽
吴道林
凤志明
LI Cong-fei TIAN Li WU Dao-lin FENG Zhi-ming(School of Electrical Engineering, Anhui Polytechnic University, Wuhu 241000, China)
出处
《陕西理工学院学报（自然科学版）》
2016年第5期57-62,68,共7页
Journal of Shananxi University of Technology:Natural Science Edition
基金
安徽省教育厅自然科学研究重点项目(KJ2014A282)
关键词
小波分解
模拟退火算法
综合诊断网络
粒子群算法
故障诊断
inverter
wavelet decomposition
simulated annealing（SA）
integrated Network（SAPSO-BP）
particle swarm optimization（PSO）
fault diagnosis
