基于变分模态分解(VMD)数据分解的多通道短期电力负荷预测模型
Multi-channel Short-Term Power Load Prediction Model Based on VMD Data Decomposition
王寅超 1蒋本建 1韩东 1杜辰坤2
作者信息
- 1. 国网上海市电力公司 经济技术研究院,上海 200030
- 2. 上海电力大学 自动化工程学院,上海 200093
- 折叠
摘要
针对电力负荷数据的随机性与不稳定性,以及传统预测模型难以准确抓取数据的局部与全局特征的问题,提出了一种基于变分模态分解(VMD)的多通道短期电力负荷预测模型(VACCLA).首先,该方法利用VMD将原始负荷数据分解为代表不同尺度的特征模态分量,以降低原始序列的不平稳度;然后,将多特征引入分解后的各模态分量,利用AdaBoost决策树算法对各个数据集进行加权处理,通过对数据特征进行放大或压缩,使模型更加注重于一些重要的数据特征;最后,使用CNN-CBAM+LSTM-ATT模型进行负荷预测.通过与多个预测模型进行对比实验,结果表明该模型相较于传统的预测方法具有更好的预测效果.
Abstract
In response to the randomness and instability of power load data,as well as the difficulty of traditional prediction models in accurately capturing local and global information of the data,this paper proposes a VMD-AdaBoost-CNN-CBAM+LSTM-ATT(Hereinafter referred to as VACCLA)multi-channel short-term power load prediction model based on variational mode decomposition(VMD).Firstly,this method utilizes VMD to decompose the original load data into feature modal components representing different scales,in order to reduce the instability of the original sequence.Then,multiple features are introduced into the decomposed modal components,and the AdaBoost decision tree optimization algorithm is used to weight and process each dataset.By amplifying or compressing the data features,the model focuses more on some important data features,Finally,use the CNN-CBAM+LSTM-ATT model for load forecasting.Through comparative experiments with multiple prediction models,the experimental results show that this model has better prediction performance compared to traditional prediction methods.
关键词
电力负荷预测/变分模态分解/AdaBoost决策树/多通道短期电力负荷预测模型(VACCLA)Key words
power load prediction/variational modal decomposition/AdaBoost decision tree/VACCLA引用本文复制引用
出版年
2024
NETL
NSTL
万方数据