An Ensemble Framework for Short-Term Load Forecasting Based on TimesNet and TCN

Chuanhui Zuo, Jialong Wang, Mingping Liu (), Suhui Deng and Qingnian Wang
Additional contact information
Chuanhui Zuo: School of Information Engineering, Nanchang University, Nanchang 330031, China
Jialong Wang: School of Information Engineering, Nanchang University, Nanchang 330031, China
Mingping Liu: School of Information Engineering, Nanchang University, Nanchang 330031, China
Suhui Deng: School of Information Engineering, Nanchang University, Nanchang 330031, China
Qingnian Wang: School of Information Engineering, Nanchang University, Nanchang 330031, China

Energies, 2023, vol. 16, issue 14, 1-17

Abstract: Accurate and efficient short-term power load forecasting is crucial for ensuring the stable operation of power systems and rational planning of electricity resources. However, power load data are often characterized by nonlinearity and instability due to external factors such as meteorological conditions and day types, making accurate load forecasting challenging. While some hybrid models can effectively capture the spatiotemporal features of power load data, they often overlook the multi-periodicity of load data, leading to suboptimal feature extraction and efficiency. In this paper, a novel hybrid framework for short-term load forecasting based on TimesNet and temporal convolutional network (TCN) is proposed. Firstly, the original load data are preprocessed to reconstruct a feature matrix. Secondly, the TimesNet transforms the one-dimensional time series into a set of two-dimensional tensors based on multiple periods, capturing dependencies within different time scales and the relationships between different time scales in power load data. Then, the temporal convolutional network is employed to further extract the temporal features and long-term dependencies of the load data, enabling a more global pattern to be obtained for temporal information. Finally, the results of load forecasting can be achieved from the fully connected layer based on the extracted features. To verify the effectiveness and generalization of the proposed model, experiments have been conducted based on the ISO-NE and Southern China datasets. Experimental results show that the proposed model greatly outperforms the long short-term memory (LSTM), TCN, TimesNet, TCN-LSTM, and TimesNet-LSTM models. The proposed model reduces the mean absolute percentage error by 20% to 43% for the ISO-NE dataset and by 10% to 31% for the Southern China dataset, respectively.

Keywords: short-term power load forecasting; multi-periodicity; TimesNet; temporal convolutional network (search for similar items in EconPapers)
JEL-codes: Q Q0 Q4 Q40 Q41 Q42 Q43 Q47 Q48 Q49 (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (2)

Downloads: (external link)
https://www.mdpi.com/1996-1073/16/14/5330/pdf (application/pdf)
https://www.mdpi.com/1996-1073/16/14/5330/ (text/html)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:gam:jeners:v:16:y:2023:i:14:p:5330-:d:1192300

Access Statistics for this article

Energies is currently edited by Ms. Agatha Cao

More articles in Energies from MDPI
Bibliographic data for series maintained by MDPI Indexing Manager ().