基于暂态负载大数据的光伏并网异常智能告警算法
Intelligent alarm algorithm for abnormal photovoltaic grid connection based on transient load big data
杨燕伟1
作者信息
- 1. 华能澜沧江新能源有限公司,云南 昆明 650051
- 折叠
摘要
光伏并网异常智能告警目前受限于静态数据,导致告警准确性低.为此,文章提出基于暂态负载大数据的异常智能告警算法.该算法通过构建暂态负载监测数据采集模型,提取中心权重向量并描述电力负荷变化,采用自回归过滤和时序特征子序列变换(Time Series Shapelet Transform,Shapelet,TSSTS)处理数据,提取时序轨迹特征,并基于卷积神经网络构建异常分级告警结构,实现深度学习并准确输出告警结果.实验结果显示,该算法的曲线下面积(Area Under the Curve,AUC)值高达 0.96,满足光伏并网异常检测要求.
Abstract
The abnormal intelligent alarm of photovoltaic grid connection is currently limited to static data,resulting in low alarm accuracy.To this end,this paper proposes an abnormal intelligent alarm algorithm based on transient load big data.By constructing a transient load monitoring data acquisition model,the algorithm extracts the center weight vector and describes the change of power load,uses self-regression filtering and time series shapelet transform,shapelet(TSSTS)to process the data,and extracts the timing trajectory characteristics.Finally,the paper constructs an abnormal classification alarm structure based on convolutional neural network,realizes deep learning and accurately outputs alarm results.Experiments show that the area under the curve(AUC)value of the algorithm is as high as 0.96,which meets the requirements of photovoltaic grid-connected anomaly detection.
关键词
暂态负载/大数据/光伏并网/特征提取/异常状态/智能告警Key words
transient load/big data/photovoltaic grid connection/feature extraction/abnormal state/intelligent alarm引用本文复制引用
出版年
2024
NETL
NSTL
万方数据