摘要
随着能源需求的增加和新型电力系统复杂性的提高,负荷的非平稳性、混沌性和非线性特征凸显。为了应对上述问题对负荷精准预测的挑战,提出一种基于多模态动态潜变量宽度学习系统的负荷多步预测方法。首先,根据负荷的非平稳特征,采用变分模态分解算法将原始负荷序列分解为多个相对平稳的模态分量,以减少非平稳性对预测的干扰。其次,针对负荷的混沌特征,基于模态分解结果,提出了一种基于动态潜变量建模的动态相空间重构方法,在相空间中提取负荷序列的动态演变趋势。最后,根据负荷的非线性特征,通过宽度学习系统挖掘并揭示负荷序列在相空间中的动态演变趋势,以完成负荷的多步预测任务。工程实际案例分析表明,提出的预测方法具有高精度的负荷预测能力。
With the increasing demand for energy and the growing complexity of modern power systems,the non-stationarity,chaotic behavior,and non-linear characteristics of loads have become more pronounced.To address the challenges posed by these issues in accurate load forecasting,a multi-modal dynamic latent variable width learning system-based multi-step load forecasting method is proposed.Firstly,to tackle the non-stationary characteristics of load data,the variational mode decomposition algorithm is employed to decompose the original load sequence into multiple relatively stationary mode components,reducing the interference of non-stationarity on forecasting.Secondly,to address the chaotic features of load data,a dynamic phase space reconstruction method based on the results of mode decomposition is proposed to extract the dynamic evolution trends of load sequences in phase space.Finally,to handle the non-linear characteristics of loads,a broad learning system is utilized to mine and reveal the dynamic evolution trends of load sequences in phase space for multi-step load forecasting tasks.The results of real-world case studies demonstrate that the proposed forecasting method exhibits high-precision load forecasting capabilities in novel modern power systems.
作者
夏蕾
张世林
李文美
谷紫文
黄纯
郭思维
王国卉
杨蕾
XIA Lei;ZHANG Shiin;LI Wenmei;GU Ziwen;HUANG Chun;GUO Siwei;WANG Guohui;YANG Lei(State Grid Henan Electric Power Company Marketing Service Center,Zhengzhou 450052,China;College of Electrical and Information Engineering,Hunan University,Changsha 410012,China)
出处
《供用电》
北大核心
2024年第1期100-108,共9页
Distribution & Utilization
基金
湖南省科技重大专项项目(2020GK1010)
国网河南省电力公司科技项目(5217X0230002)。
关键词
负荷预测
变分模态分解
动态潜变量建模
宽学习系统
相空间重构
混沌时间序列
load prediction
variational mode decomposition
dynamic latent variable modeling
broad learning system
phase space reconstruction
chaotic time series
