基于小波神经网络的电力系统振荡和故障识别

目前距离保护中的振荡闭锁元件都不同程度地导致振荡中故障的延时及无选择切除，基于此，文中综合小波变换以及神经网络的突出优点，构建了一种新型的小波神经网络模型，并给出了其相应的算法，以此小波神经网络实现了高压输电线路距离保护中基于暂态信号的系统振荡闭锁元件。理论分析及大量EMTP仿真实验表明：充分训练学习后的小波网络能够正确、快速地识别系统振荡和各种故障情况，即使是系统振荡时最不利情况下的线路轻微故障，亦能获得比较满意的结果，且此方法具有较高的可靠性。

WAVELET NEURAL NETWORKS BASTED RECOGNITION OF SWING AND FAULT IN POWER SYSTEM

All of the existing power swing blocking elements would cause, at different extent, delayed and blind elimination during the power swing. This paper presents a new type of wavelet neural networks (WNN) model with the integration of the outstanding characteristics of Wavelet transform (WT) and Neural Networks (NN), and its corresponding algorithm. Based on the WNN, a new principle for power swing block using transient signal could be designed in the distance protection devices. Theoretical analysis and lots of EMTP simulation results show that WNN after enough learning can quickly and correctly recognize the fault during power swing. Even under the unfavorable conditions, satisfactory results can be achieved. And the method has many advantages such as fast computation and response, high reliability etc.