基于反向传播神经网络的失流故障智能识别研究

为解决传统方法识别主网电流失流故障准确率低的问题，本文提出了一种基于反向传播神经网络的失流故障智能识别方法。本文利用反向传播的神经网络算法，通过梯度下降的方式反向修正各层权值，使网络输出误差达到可以接受的程度，从而达到对失流故障识别具有很好的自学习自适应能力的目的。首先根据对主网失流故障特征的研究，将失流故障分为持续失流与断续失流两种，构造对应指标，综合所有指标构建失流故障特征提取体系，最后建立反向神经网络来拟合失流故障提取体系。对数据进行识别，建立专家样本库，利用反向神经网络进行离线训练，训练完成后固定权值用于失流故障识别，从而准确输出失流故障事件。经实例验证，所提方法在识别准确率和识别效率优于一般分类识别方法，可实现失流故障的就地识别。

Intelligent Identification Research of Current-lossing Fault Based on BP Neural Network

Aiming at the problem of low accuracy of current-lossing identification in main network by traditional methods, an intelligent identification method of current-lossing based on Back Propagation (BP) neural network is proposed. In this paper, make use of back-propagation neural network algorithm, through the mode of gradient descent to reverse the weights of each layer, the network output error is acceptable, and it has good self-learning ability for fault identification of current-lossing. According to the research of current-lossing fault in this paper, firstly construct indicators of continuous current-lossing and discontinuous current-lossing, build up the fault feature extraction system based on all indicators, finally BP network is established to fit current-lossing extraction system, Identify the data, then Set up the expert samples, and train BP neural network off-line, fix weight value after training. Through the test by example, this model is superior to the general classification model and has high discrimination of current-lossing fault recognition. It can provide a forceful basis for solving the problem.