【摘要】 提出了一种基于最小二乘递推法(RLS)的正交基神经网络算法来分析电力系统谐波参数.该方法根据谐波分析的特点,采用RLS训练神经网络权值,有效地避免了梯度下降法存在局部极小的问题,并且对降低噪声影响有显著作用.电力系统谐波分析的仿真结果表明,该算法经过一次神经网络训练即可获得各次谐波高精度的幅值和相位.更多还原

【Abstract】 An algorithm of neural network with orthogonal basis functions based on recursion least square(RLS) was presented and used to analyze the harmonics of power system.The main idea was to use RLS algorithm to train the weights of neural network according to the harmonics characteristic.Therefore,the local least problem on grads descent method was effectively avoided,and the noise was evidently filtered.To validate the validity of the algorithm,simulation examples of harmonic analysis by the use of the presented approach were given.Simulation results have shown that the exact amplitudes and phases of the harmonics were obtained by training neural network only one time,so it will be very valuable in power system harmonic analysis.更多还原