基于Seq2point-ASPP的非侵入式负荷分解算法
Non-invasive load decomposition algorithm based on Seq2point-ASPP
王俊泽 1郭晓雪 1黎赛1
作者信息
- 1. 西安工程大学 电子信息学院 西安 710048
- 折叠
摘要
随着电网升级,我国配电网迅速发展.非侵入式负荷监测作为智能电网中的先进测量技术,能准确服务电力系统用户并促进用户形成良好用电习惯,实现互利.现有非侵入式负荷分解技术存在计算量大、训练效率低等挑战,本文提出一种基于序列到点和空洞空间金字塔池化的方法.通过结合2D卷积对Seq2point模型优化,将参数量从3000万降至180万,显著提升训练效率.此外,ASPP模块的引入能更好地捕获不同尺度的特征,优化特征提取.模拟实验表明,Seq2point-ASPP方法在负荷分解效果上优于传统方法,平均MAE可达5.17,平均SAE可达0.051,具有快速计算和高实用价值的优点.
Abstract
With the upgrading of power grid,China's distribution network develops rapidly.As an advanced measurement technology in smart grid,non-invasive load monitoring can accurately serve power system users and promote good usage habits to achieve mutual benefit.Although the existing non-invasive load decomposition techniques have many challenges in computation and low training efficiency,this paper proposes a method based on sequence-to-point and void space pyramid pool.By combining 2D convolution to optimize the Seq2point model,the number of parameters was reduced from 30 million to 1.8 million,significantly improving the training efficiency.In addition,the introduction of ASPP module can better capture features of different scales and optimize feature extraction.The simulation results show that the Seq2point-ASPP method is superior to the traditional method in load decomposition,with an average MAE of 5.17 and an average SAE of 0.051.It has the advantages of fast calculation and high practical value.
关键词
非侵入式负荷分解/金字塔池化/序列到点/深度学习Key words
non-invasive load decomposition/pyramid pooling/sequence-to-point/deep learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据