基于概率神经网络和双小波的电能质量扰动自动识别

对电能质量(PQ)扰动的自动识别是找出引起PQ问题根本原因的前提。提出了一种基于概率神经网络PNN(Probabilistic Neural Networks)和db10,db1双小波的PQ扰动自动识别方法。首先,利用db10小波对信号进行分解,将各层小波变换系数的能量和第1高频层模极大值情况作为PNN的输入矢量,判断扰动类型;然后,对信号进行傅里叶变换以检测信号中是否含有谐波;最后,对判断存在电压下降的信号进行db1小波分解,根据其低频层的模值区分电压下陷和电压中断信号。测试结果表明,该方法提高了识别正确率,且实现简单,能有效检测幅值较小的谐波。

Auto recognition of power quality disturbance based on probabilistic neural networks and double wavelet

Auto recognition of PQ(Power Quality) disturbance is the premise of finding out the ultimate cause of PQ problems. A method based on PNN(Probabilistic Neural Networks) and double wavelet of db10 and db1 for auto recognition of PQ disturbance is proposed . The signal is decomposed by db10 wavelet,the energy of wavelet transform coefficients at different levers and the status of the first layer high-frequency decomposition coefficients module maximum are severed as the input vectors of PNN for disturbance type judgment . Fast Fourier transform is applied to the signal to detect its harmonics . The signal with voltage drop is decomposed by db1 wavelet,and the voltage sag is distinguished from voltage interrupt signal by the module of the low frequency coefficients. Test results show that the correct recognition ratio is improved. The proposed method is easy to implement and detect the harmonic with low magnitude effectively.