【摘要】 提出了一种微机变压器保护的新方法。利用小波分析的信号奇异性检测特性,建立起信号突变量与模局部极大值的一一对应关系,提取所需要的状态特征;利用小波分析的时频分解滤波特性,减少故障暂态过程中非周期分量和高次谐波对故障方向判别的影响。在此基础上,提出并建立三层前向神经网络模型,用来实现变压器的微机保护。理论分析和EMTP仿真试验表明,该方法可以有效地提高保护的灵敏性、选择性和速动性,而且不受CT饱和的影响。更多还原

【Abstract】 This paper presents a new method of computerized transformer protection.Using wavelets analyzed signal fantasticality inspection characteristic,it sets up the corresponding relationship between values of signal discontinuity and its modular partial maxima, and picks up state characteristics that needed.Using the characteristic of its time-frequency decomposing filter, it reduces the influence of inperiodic components and higher harmonic to fault direction identification in transient fault. Based on these,a three-layer feed-forward neural network model is presented and built in order to realize the computer-based transformer protection.Theoretical analysis and EMPT simulation test show that the method can effectively improve sensitivity and selectivity of conventional numerical protection principles and criteria without the effect of CT saturation.更多还原