光伏发电机组隐性故障快速序列化预警技术  

Rapid serialization and early warning technology for hidden faults in photovoltaic generation units

作　　者：卜乐 雷阳 BO Le;LEI Yang(State Grid Shaanxi Electric Power Co.,Ltd.Marketing Service Center,Xi’an 710199,China)

机构地区：[1]国网陕西省电力有限公司营销服务中心(计量中心),陕西西安710199

出　　处：《电子设计工程》2025年第5期142-146,共5页Electronic Design Engineering

基　　金：国网陕西省电力有限公司科技项目(5226KY2360)。

摘　　要：隐性故障会造成光伏发电机组运行失稳,为了有效预警光伏发电机组隐性故障,提出了光伏发电机组隐性故障快速序列化预警技术。通过BiLSTM反向传播,解决数据库故障长文本梯度弥散问题。利用相邻标签关系,获取一个全局最优标签序列。根据标注序列在所有序列集合中的softmax概率得分,获取处理后的序列化故障数据。计算隐性故障支持度,缩短隐性故障序列化挖掘时间。设置挖掘关联规则,使用频繁挖掘方法挖掘序列化隐性故障。根据站域保护隐性故障分析结果,快速预警隐性故障。由实验结果可知,使用所研究技术预警的故障时间与实际故障发生时间一致,预警的电压波动幅度仅与实际情况存在0.03 pu的偏差,具有精准预警效果。Hidden faults can cause unstable operation of photovoltaic generation units.In order to effectively warn of hidden faults in photovoltaic generation units,a rapid serialization warning technology for hidden faults in photovoltaic generation units has been proposed.Solve the problem of long text gradient dispersion in database faults through BiLSTM backpropagation.Utilize adjacent label relationships to obtain a globally optimal label sequence.Obtain processed serialized fault data based on the softmax probability score of the annotated sequence in all sequence sets.Calculate implicit fault support and shorten the mining time for implicit fault serialization.Set mining association rules and use frequent mining methods to mine hidden faults in serialization.Based on the analysis results of hidden faults in site protection,quickly alert hidden faults.From the experimental results,it can be seen that the fault time of the studied technology warning is consistent with the actual fault occurrence time,and the voltage fluctuation amplitude of the warning only deviates by 0.03 pu from the actual situation,which has a precise warning effect.

关 键 词：光伏发电机组 隐性故障 快速序列化 故障预警 

分 类 号：TN92[电子电信—通信与信息系统]