基于S变换与PNN的电能质量多扰动检测

Power quality disturbance detection based on S-transform and PNN

针对电能质量检测与分类需求,提出了一种基于S变换与概率神经网络的电能质量扰动检测和分类方法,应用S变换对电能质量扰动样本信号进行时频分析,提取信号的特征量,利用获得的特征量训练概率神经网络,并进行分类。仿真实验证明基于S变换与概率神经网络融合的电能质量多扰动分类方法训练速度快、分类准确度高,在训练样本数少、噪声影响大和多扰动信号并存时分类识别效果好。在此基础上研制了基于虚拟仪器的电能质量扰动检测系统,给出了系统构成与工作流程,现场试验验证了系统的准确性。

A novel detection and classification method of power quality disturbances based on S-transform and probabilistic neural network （PNN） is proposed. S-transform is applied to perform time-frequency analysis on the power quality disturbance samples, from whose results the features of the samples are extracted. These features are then used to train a PNN for disturbance classification. Simulation results show that this method has faster training speed and relatively high classification accuracy. The proposed method can also give good classification for small training set with high level noises and multiple types of disturbances. Furthermore, a power quality disturbance detection system based on virtual instrument was developed. The system structure and work flow are given. Experimental results of on-spot operation verify the veracity of the system.