基于神经网络建模的零序直流原理测距研究

针对电力电缆铺设隐蔽,发生故障时不易准确定位故障点的问题,研究了零序直流保护原理和人工神经网络技术在小电流接地系统单相接地故障测距中的应用。利用直流信号不受电网电容影响的特性,建立检测源与接地网电路模型。根据电路模型,建立其数学模型,由于模型中的一些参数复杂,则考虑采用神经网络对其未知量进行辨识,并将辨识好的数学模型应用于故障测距中。根据实验数据,对两种模型进行分析,结果表明利用神经网络辨识参数的方法能够准确、可靠地实现故障测距。

Study on Fault Location Based on Neural Network Modelling and Zero-sequence DC Principle

In view of the difficulty in fault location for concealed power cables, this paper studies the application of zerosequence DC protection principle and artificial neural network （ANN） technology to location of single-phase ground fault of small-current grounding system. As the DC signal is immune to the effect of power grid capacitance, a circuit model of detecting source and grounding grid is established. According to the circuit model, the mathematical model of the fault location system is established. Due to the complexity of some parameters in the model, a method which uses ANN to identify the unknown parameter is proposed. The identified mathematical model is applied in the cable fault location. The two models are analyzed based on the experimental data. It is demonstrated that the ANN method is accurate and reliable in fault location.