基于自定义深度学习网络双馈风机接入弱电网故障识别

Fault Recognition under Weak Grid Condition based on Doubly Fed Induction Generator With Customized Deep Learning Network

针对双馈风机接入弱交流电网的故障检测问题,提出一种自定义深度学习网络框架下的故障识别和算法。首先采集实际系统的运行数据,通过归一化操作将其转化为零均值数据,然后用自定义深度学习网络训练数据集,形成预训练网络。同时模拟异常数据,输入至此前的预训练网络中,输出预测数据。根据预测数据和模拟异常数据计算出阈值,当预测数据和模拟异常数据差值绝对值超出阈值后,判断系统数据异常。仿真结果表明,所提方法能够比较准确地解决故障识别问题,可以被用于含双馈风机弱电网中故障检测。

In view of the problem of fault detection under the condition of doubly-fed wind turbine access to the weak AC grid at present,a fault recognition and detection algorithm under the framework of customized deep learning network was proposed.Firstly,the actual system data was used and normalized to zero-mean standard data,and then the data set was trained by customized deep learning network to form a pre-training network.At the same time,abnormal data was simulated and input to the previously pre-trained network to output prediction data.The threshold was calculated according to the prediction data and simulated abnormal data.When the absolute value of the difference between the prediction data and the simulated abnormal data exceeds the threshold,the system data was judged to be abnormal.The simulation results showed that the proposed method can solve the fault recognition problem more accurately and meet the requirements of fault detection in a weak grid containing doubly-fed wind turbines.