A multi-objective optimization-based ensemble neural network wind speed prediction model
With the depletion of fossil fuels, wind energy, as a renewable resource, is being widely developed worldwide. Wind speed prediction plays a crucial role in the efficient utilization of wind energy. However, most existing short-term wind speed forecasting methods rely solely on historical wind speed...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | International Journal of Electrical Power & Energy Systems |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S0142061525003813 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | With the depletion of fossil fuels, wind energy, as a renewable resource, is being widely developed worldwide. Wind speed prediction plays a crucial role in the efficient utilization of wind energy. However, most existing short-term wind speed forecasting methods rely solely on historical wind speed sequences, neglecting other influential factors. Additionally, there is a lack of effective ensemble models capable of integrating the strengths of various neural networks, and current hyperparameter optimization strategies often fail to simultaneously address the issues of overfitting and underfitting. To overcome these limitations, this paper proposes a short-term multi-step wind speed prediction model that incorporates six meteorological factors as input features. The model integrates six distinct neural network architectures using eXtreme Gradient Boosting (XGBoost) as the meta-learner. To optimize the hyperparameters of XGBoost, we introduce a novel algorithm named NS-ADPOA, which adopts a bi-objective optimization strategy targeting both Mean Squared Error and model complexity. Built upon the NSGA-II framework, NS-ADPOA enhances offspring generation by leveraging a probabilistic error-driven fusion of Particle Swarm Optimization (PSO) and Whale Optimization Algorithm (WOA), combining their strengths in local and global search, respectively. We conduct experiments on four datasets representing diverse wind conditions. The results from four different experimental settings demonstrate that the proposed model consistently achieves the best prediction accuracy across all regions, with maximum improvements of up to 96.5% in MSE, 81.3% in Root Mean Squared Error (RMSE), and 84.5% in Mean Absolute Error (MAE) compared to baseline models. These findings confirm that the proposed approach effectively integrates the complementary advantages of multiple neural networks. Moreover, the NS-ADPOA algorithm successfully balances model complexity and accuracy, enabling the XGBoost model to mitigate both underfitting and overfitting, and achieving high accuracy in multi-step wind speed prediction. |
|---|---|
| ISSN: | 0142-0615 |