考虑电缆故障情况下风电场电气内部故障诊断方法

Diagnosis Method for Electrical Internal Faults in Wind Farms Considering Cable Faults

针对传统的风电场维护方案滞后性强、运行成本高,提出了一种考虑电缆故障的风电场电气内部故障诊断方案。先利用傅里叶分解算法对风电场故障信号进行初步处理,得到了内秉带函数和残余信号,保障故障信号平稳,便于后续识别;然后对改进图卷积神经网络进行故障识别和诊断,设计了故障信息串联模型,并对故障进行分类,结合代价敏感学习调整分类结果,保障系统对样本较少的风电场仍能够实现故障诊断。试验结果表明,所提方案能有效实现对风电场电气内部故障的识别和诊断,抗干扰能力强,不易受噪声影响,故障识别准确率达到96%。

A wind farm electrical internal fault diagnosis scheme considering cable faults was proposed to address the strong lag and high operating costs of traditional wind farm maintenance schemes.Firstly,the Fourier decomposition algorithm was used to preliminarily process the fault signal of the wind farm,obtaining the Fourier intrinsic band functions(FIBFs)and residual signal,ensuring the stability of the fault signal and facilitating subsequent identification;the graph convolutional neural network was improved for fault identification and diagnosis,and a series model of fault information was designed to classify the faults.Combined with cost sensitive learning,the classification results were adjusted to ensure that the system can still achieve fault diagnosis for wind farms with less samples.The experimental results show that the proposed scheme can effectively identify and diagnose electrical internal faults in wind farms with strong anti-interference ability,and is not easily affected by noise with a fault identification accuracy of 96%.