User abnormal load identification method based on edge computing
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
万方数据
针对未知家电负荷背景下基于智能电表采样数据进行异常负荷识别问题,以电动车入户充电负荷为出发点,提出了一种基于边缘计算的用户异常负荷识别方法.首先通过Boruta-SHAP算法对非侵入式负荷数据的14种特征进行排序筛选,得到在秒级负荷数据下的辨识效果最佳的特征子集;然后采用改进的非平行支持向量机(v-non parallel support vector ma-chine,v-NPSVM)模型进行异常负荷识别模型的训练;最后结合边缘计算技术将算法部署到边缘计算平台上,实现对典型电动车充电负荷的识别.实验基于低压台区中智能电表获取的真实负荷数据进行验证,并进一步对数据进行降频处理以验证更低频数据源下方法的有效性,实验结果表明针对降频后的异常负荷识别的正确辨识率仍在90%以上,证明了在未知家电负荷背景下方法具有较好的适用性和准确性.
In order to solve the problem of abnormal load identification based on smart meter sampling data in the context of unknown household appliance load,this paper proposes a user abnormal load identification method based on edge computing based on the charging load of electric vehicles.Firstly,the Boruta-SHAP algorithm was used to sort and screen 14 features of the non-intrusive load data,and the subset of features with the best identification effect under the second-level load data was obtained.Then,the v-non parallel support vector machine(v-NPSVM)model was used to train the abnormal load recognition model.Finally,the algorithm is deployed on the edge computing platform combined with edge computing technology to realize the identification of typical electric vehicle charging loads.The experimental results show that the correct identification rate of abnormal load identification after frequency reduction is still more than 90%,which proves that the proposed method has good applicability and accuracy in the context of unknown household appliance load.
non-intrusive load identificationedge computingv-NPSVMabnormal loaduser charging for electric vehi-cles
NETL
NSTL
万方数据
