基于多任务辅助学习的配网低电压成因分析

Analysis on the causes of low voltage in distribution network based on multi-task assisted learning

当前配网低电压愈发严重,已经严重影响居民的日常生活,而维护工单反馈模糊,难以准确定位其原因。为了准确定位配网低电压的成因,提出一种基于多任务辅助学习的配网低电压成因分析模型。首先,获取低电压用户96点电流、电压等原始数据,并实现原始数据的预处理;其次,利用双向门控循环单元神经网络(bidirectional gated recurrent unit,BiGRU)挖掘数据的深度特征;最后,将引发配网低电压的主成因分析设置为主任务,子成因的分析作为相关辅助任务,利用相关辅助任务强化数据中隐藏特征学习,为主任务提供额外的监督信息,并采用多任务联合训练方式训练主成因分析模型,协助模型学习到更具鲁棒性的特征表示,提高配网低电压成因分析的准确率。实验结果表明,提出的基于多任务辅助学习的配网低电压成因分析模型具有较好的分析定位能力,最终分类准确率可达95.58%。

The current low voltage(LV)in the distribution network is becoming more and more serious,which seriously affects the daily life of residents,and the feedback of maintenance work orders is vague and unable to accurately locate the cause.In order to solve this problem,a LV cause analysis model based on multi-task assisted learning is proposed in the paper.Firstly,raw data such as current and voltage at 96 points of the LV users are obtained,and the pre-processing of raw data is achieved.Secondly,the deep features of the data are mined by using bidirectional gated recurrent unit(BiGRU)neural network,at the same time,the analysis of the main cause of LV is set as the main task,and the analysis of sub-causes is set as the auxiliary tasks,and which are used to strengthen the learning of hidden features in the data and provide additional supervisory information for the main task.Multi-task joint training is used to train the main cause analysis model,assist the model to learn more robust feature representations and improve the accuracy of LV cause analysis.The experimental results show that the LV causal analysis model based on multi-task assisted learning proposed in this paper has better analysis and localization ability,and the final classification accuracy can reach 95.58%.