.png){kind=icon}
返回
Attention-based temporal-spatial convolutional network for ultra-short-term load forecasting
.png){kind=icon}
PRE
AI
DOI：10.1016/j.epsr.2023.109329.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Load forecasting plays a critical role in energy dispatch and in maintaining the efficient, stable, safe, and economical operation of energy systems. Although the current deep learning-based hybrid methods have achieved good results, their huge model sizes can lead to colossal resource occupations in practical deployments. This paper proposes a novel powerful attention-based temporal-spatial convolutional network (ACN), reaching the top forecasting accuracy with fewer parameters for ultra-short-term load forecasting. The proposed ds-inception and 1D convolutional block attention module (CBAM) structures ensure the feature learning capability of the basic ACN, allowing the ACN-based forecasting model to achieve better results with less layer stacking. Meanwhile, a multi-head attention mechanism is adopted as the regression module to strengthen the long-term dependencies of the stacked ACN. The robustness and prediction accuracy were proven on real-world datasets from China and England. Compared to the best state-of-the-art models, an average reduction in mean absolute percentage error (MAPE) of 7.76% and root mean squared error (RMSE) of 2.37% were achieved by ACN, with parameter numbers similar to the original temporal convolutional network.
Keyword:
Energy internet
Ultra-short-term load forecasting
Temporal convolutional network
Attention mechanism
期刊
IF:
4.2
论文数:
1.2W
被引数:
2.2W
机构
暂无机构信息
引用论文
Short-Term Load Forecasting for Industrial Customers Based on TCN-LightGBM基于tcn-lightgbm的工业客户短期负荷预测
PREAI
分享收藏Caw’s Walking State Recognition Based on Accelerometers and Gyroscopes Installed on Ear-Tags and Collar-Tags基于安装在耳标和项圈上的加速度计和陀螺仪的Caw步行状态识别
PREAI
分享收藏Global Learnable Pooling With Enhancing Distinctive Feature for Image Classification.png)
IEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAI分享收藏A Survey on Energy Internet: Architecture, Approach, and Emerging Technologies能源互联网: 架构、方法和新兴技术
PREAI
分享收藏Synthesis of n-type semiconducting diamond film using diphosphorus pentaoxide as the doping source以五氧化二磷为掺杂源合成n型半导体金刚石膜
PREAI
分享收藏A Novel CNN-GRU-Based Hybrid Approach for Short-Term Residential Load Forecasting基于cnn-gru的短期住宅负荷预测新方法IEEE ACCESS
OAAI
IEEE ACCESSIF3.6
OAAI分享收藏
查看更多内容