基于级联宽度学习与麻雀算法的非侵入式负荷分解方法
Non-intrusive load disaggregation method based on cascade broad learning and sparrow algorithm
白星振 1康家豪 2尚继伟 1郝春蕾 1王雪梅3
作者信息
- 1. 山东科技大学 电气与自动化工程学院,山东 青岛 266590
- 2. 国网山东省电力公司 汶上县供电公司,山东 济宁 272500
- 3. 山东科技大学 工程实训中心,山东 青岛 266590
- 折叠
摘要
深度学习被广泛应用于非侵入式负荷分解中,其分解精度高但存在网络结构复杂、训练过程极度耗时等问题,并且对计算资源有一定要求,难以与嵌入式设备集成使用.对此,面向低频数据,提出一种基于级联宽度学习与麻雀算法的非侵入式负荷分解方法.首先,改进宽度学习特征节点的连接方式,构建各目标设备的级联宽度学习负荷分解网络.然后,通过麻雀搜索算法确定各目标设备分解网络的最优特征节点和增强节点数,实现负荷的高效分解.最后,基于实际数据集UK-DALE进行了仿真实验,通过与常用的非侵入式负荷分解方法进行比较,验证了所提方法的优越性.
Abstract
Deep learning has been widely used in non-intrusive load disaggregation.Despite its high disaggregation accuracy,its problems such as complex network structure,extremely time-consuming training process,and some requirements for computational resources make it difficult to integrate with embedded equipment.To address these problems,a non-intrusive load disaggregation method based on cascaded broad learning and sparrow optimization was proposed for low-frequency data.Firstly,the connection method of the broad learning feature nodes was improved and the cascade broad learning load disaggregation network of each target device was constructed.Then,the optimal number of feature nodes and enhanced nodes of the disaggregation network of each target device was determined by the sparrow optimization algorithm to realize the efficient disaggregation of non-intrusive load.Finally,simulation experiments based on the real dataset UK-DALE were carried out,and the superiority of the proposed method was verified by comparing it with the commonly-used non-intrusive load disaggregation methods.
关键词
非侵入式/负荷分解/宽度学习/麻雀算法/特征节点级联Key words
non-intrusive/load disaggregation/broad learning/sparrow algorithm/cascade of feature nodes引用本文复制引用
出版年
2024
NETL
NSTL
万方数据