基于人工神经网络的AT单线供电系统故障测距研究

Study on Artificial Neural Network-based Fault Location of AT Single Line Power Supply System

AT单线牵引供电系统由于AT变压器、避雷线圈的接入以及供电线与接触网的材质不同等问题,传统测距方法未能对短路数据进行科学系统分析,故障发生时精确定位仍比较困难。为解决上述问题,本文引入了人工神经网络自动测距方法,首先推导出AT单线牵引供电系统故障量与故障点距离的关系;通过搭建人工神经网络模型,对不同位置、不同短路类型的数据进行训练,训练成功的网络对随机故障位置进行预测,精度满足工程要求;最后搭建相应的试验系统,对理论分析的正确性进行验证。试验结果与理论分析具有较好的一致性,证明了所提方法与结论的正确性,为进一步研究AT单线牵引供电系统故障定位提供了一种新的有益参考。

Due to the problems of access of AT transformer and lightning coils as well as the problem that the materials of power supply feeders and OCS are different for AT single line traction power supply system,the traditional fault location methods fail to analyze the short circuit data scientifically and systematically,and the accurate fault location is still difficult while the fault occurs.In order to solve the above-mentioned problems,an automatic fault location method based on artificial neural network is introduced for deducting the distance relations between fault quantity and fault point of AT single line traction power supply system at first;then executing the training on data of different locations and of different types of short circuits on the basis the established artificial neural network model afterwards;establishing the artificial neural network model for execution of training of data from different places and data of different types of short circuits,executing the prediction on random fault location by use of successfully trained network.Its accuracy conforms to the engineering requirements;establish finally the related test system for verification of correctness of the theoretical analysis.The test results and theoretical analysis have better consistency,verifies the correctness of the proposed methods and results,providing a new and beneficial reference for further studies on fault location of AT single traction power supply system.