CAJ | 학술논문

在选取BP神经网络对谐波进行分析时，考虑到BP网络存在的缺点和不足，对BP网络进行改进。通过把FFT和改进的BP网络结合起来，实现对谐波的实时分析。该方法先对采样信号进行FFT运算实现预处理，得到谐波个数和谐波次数；然后根据谐波的个数来确定神经元的个数，通过谐波次数设定神经网络参数迭代的初始值；最后对改进的神经网络进行训练，可以实现谐波的精确分析。所提出的BP改进算法有效地提高了谐波分析的精度，实时性也得到了改善。
재선취BP신경망락대해파진행분석시，고필도BP망락존재적결점화불족，대BP망락진행개진。통과파FFT화개진적BP망락결합기래，실현대해파적실시분석。해방법선대채양신호진행FFT운산실현예처리，득도해파개수화해파차수；연후근거해파적개수래학정신경원적개수，통과해파차수설정신경망락삼수질대적초시치；최후대개진적신경망락진행훈련，가이실현해파적정학분석。소제출적BP개진산법유효지제고료해파분석적정도，실시성야득도료개선。
Considering the disadvantages of BP network, a BP neural network is improved to analyze harmonics. The FFT and the improved BP neural network are combined effectively to analyze the harmonic in real time. It first?ly pre?processes the sampled data signal by analyzing the signal with FFT algorithm, so as to get the number and or?der of harmonic waves, then determines the number of neurons according to the number of harmonic waves and sets the initial value of neural network parameter iteration, and finally trains the improved neural network, thereby reali?zing precise harmonic analysis. Simulation results show the improved BP neural network can improve the accuracy and the real?time performance of harmonic analysis.