Wind Farm Meteorological Prediction Model Based on Frequency Domain Feature Extraction Fusion Mechanism
Given the increasing significance of wind energy in the worldwide energy composition, the effective administration of wind farms has emerged as a crucial element in guaranteeing the reliability of the energy provision. Precise forecasting of meteorological conditions is essential for optimizing wind...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10649551/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Given the increasing significance of wind energy in the worldwide energy composition, the effective administration of wind farms has emerged as a crucial element in guaranteeing the reliability of the energy provision. Precise forecasting of meteorological conditions is essential for optimizing wind farm operations in this scenario. However, the current prediction model has certain limitations, namely the problem of inaccurate prediction for a long period and the problem of insufficient ability to extract frequency domain information. Among them, we found that frequency domain features are the key to affecting the above problems. Therefore, this study aims to enhance the ability of existing models to capture frequency domain feature information by introducing an innovative frequency domain feature extraction fusion mechanism, thereby significantly improving the accuracy of meteorological condition forecasting. Specifically, this study addresses the issue of long-term dependency in the model by incorporating the Frequency Enhanced Attention module. This integration improves the model’s ability to reliably forecast wind farm meteorological conditions over extended periods. Furthermore, this study utilizes the LoRA (Low-Rank Adaptation) fine-tuning strategy to improve the model’s generalization capacity across various wind power circumstances. In the experimental part, we verify the model’s high-precision forecasting ability on coarse-grained seasonal changes. At the same time, we have extensively verified the model in four different regions, where our model outperformed existing mainstream models (Autoformer, Informer, and Transformer) by an average of 28.9%, 22.7%, and 27.0% respectively, demonstrating the broad applicability of our model. The model prediction results are better than the current mainstream models in critical indicators, proving the model’s effectiveness. These experiments validate our model’s efficacy and offer novel technical resources for forecasting meteorological conditions and managing energy in wind farms. This is highly significant in advancing the efficient utilization of wind energy and promoting the sustainable development of the wind power industry. |
|---|---|
| ISSN: | 2169-3536 |