PV and Load Forecasting for Photovoltaic Communities Based on a Hybrid Model of LSTM and Transformer

This paper investigates the PV and load forecasting problem for PV communities. A construction of a PV and load forecasting model based on a hybrid model of LSTM and Transformer is proposed. The multi-attention mechanism of Transformer is added to the traditional neural network LSTM model. This hybr...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:IEEE International Conference on Industrial Technology (Online) s. 1 - 6
Hlavní autori: Cen, Jinglong, Zhang, Xing, Wang, Qingyi, Xiao, Zhongyuan, Xu, Chang, Li, Xiang
Médium: Konferenčný príspevok..
Jazyk:English
Vydavateľské údaje: IEEE 26.03.2025
Predmet:
Attention mechanisms
Correlation
Data models
Load forecasting
Load modeling
Long short term memory
multiple attention mechanisms
Neural networks
prediction
Predictive models
time Series
Time series analysis
Transformers
ISSN:2643-2978
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:This paper investigates the PV and load forecasting problem for PV communities. A construction of a PV and load forecasting model based on a hybrid model of LSTM and Transformer is proposed. The multi-attention mechanism of Transformer is added to the traditional neural network LSTM model. This hybrid model optimizes the original data according to the PV and load time series characteristics. The weighted mean square error method is used to design the loss function. In order to prevent model overfitting and improve stability, Adam optimizer and regularization coefficient is introduced. In order to verify the superiority of the model, machine learning and traditional neural network method comparison experiments are carried out in this paper. The experimental results show that the model designed in this paper has the smallest error between the predicted value and the true value. It shows that the model with the introduction of Transformer's multi-attention mechanism in LSTM model can improve the prediction accuracy.
ISSN:2643-2978
DOI:10.1109/ICIT63637.2025.10965325