Cumulative Load Forecasting in Microgrid Clusters Using Enhanced-Gated Recurrent Unit Model
With the rise of microgrids integrated with virtual power plants (VPPs), the importance of coordinated dispatch among microgrid clusters in managing distributed energy resources has significantly increased. Accurately understanding the load trends of microgrid clusters facilitates the operation of d...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10811929/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With the rise of microgrids integrated with virtual power plants (VPPs), the importance of coordinated dispatch among microgrid clusters in managing distributed energy resources has significantly increased. Accurately understanding the load trends of microgrid clusters facilitates the operation of decentralized energy systems and enhances scheduling flexibility. Traditional day-ahead load forecasting models typically focus on specific load types. However, the diverse load structures within individual microgrids hinder these models from comprehensively capturing the dependencies and coupled characteristics among different microgrids, resulting in limited forecasting accuracy. To solve the disadvantage of existing models, this study proposes an enhanced Gated Recurrent Unit (GRU) model based on the Layer-wise Adaptive Moments optimizer for Batch training (LAMB). By adjusting the layer-wise adaptivity during gradient updates, the proposed model improves the accuracy and stability of time series forecasting for complex load structures in microgrid clusters. The effectiveness of the model is validated using aggregated load datasets from three microgrids in Tetouan, Morocco. Numerical results indicate that the LAMB-GRU model achieves a coefficient of determination (<inline-formula> <tex-math notation="LaTeX">$R^{2}$ </tex-math></inline-formula>) consistently ranging from 0.978 to 0.985, with the lowest root mean square error (RMSE) and mean absolute percentage error (MAPE) across four scenarios. Furthermore, the p-values significantly exceed the significance level, demonstrating superior performance compared to traditional day-ahead forecasting methods. The proposed model effectively reduces prediction errors and enhances forecasting stability under complex, nonlinear aggregated load conditions. |
|---|---|
| ISSN: | 2169-3536 |