【摘要】 提出了一种基于人工神经网络(artificial neural network, ANN)的高压直流系统故障诊断方法。首先根据某直流输电系统换流站近3年来实测故障数据，分析了直流输电系统换流站主要故障类型，将其归纳划分为交流故障、直流故障、换流阀故障和换相失败4种类型。然后根据场站实测数据建立故障数据集，收集整理4类故障在15个周波内具有代表性特征意义的11个通道故障特征；在此基础上，将故障样本随机划分为训练集和测试集，并更新模型网络参数，完成模型训练。最后设置算例，使用测试集对诊断模型进行测试，并设置支持向量机(support vector machine, SVM)和Naive Bayes作为对比。结果表明，所提出的方法对高压直流系统的故障诊断准确率可以达到91%以上，具有较高准确率，适用于高压直流系统的故障识别和诊断。更多还原

【Abstract】 It proposes a fault diagnosis method for high-voltage direct current(HVDC) systems based on artificial neural networks(ANNs).Firstly, the main fault types of direct current transmission system converter stations are analyzed based on the measured fault data of a direct current transmission system converter station in the past 3 years, and they are classified into 4 types: AC fault, DC fault, converter valve fault, and commutation failure.Then, based on the measured data from the site, a fault dataset is established, collecting and organizing 11 channel fault features that are representative and significant within 15 cycles for the 4 types of faults.On this basis, the fault samples are randomly divided into training and testing sets, and the model network parameters are updated to complete model training.Finally, it sets up a calculation example, and testes the diagnosis model using the testing set.Support vector machine(SVM) and Naive Bayes are set up as comparisons.The results show that the proposed method can achieve an accuracy rate of over 91% for fault diagnosis of HVDC systems, with high accuracy and strong robustness, and is suitable for fault identification and diagnosis of HVDC systems.更多还原