Trans GAN-WT: A Feature Extraction and Interactive Learning-Based Anomaly Detection Model for Wind Turbine Time Series Data
Trans GAN-WT: A Feature Extraction and Interactive Learning-Based Anomaly Detection Model for Wind Turbine Time Series Data
Authors: Jingzhe Kang Date: 2026-06-02 Paper ID: openalex:2606.03112
Summary
TransGAN-WT is an anomaly detection framework designed for wind turbine operational data that integrates Transformer-based feature extraction with generative adversarial networks. The model improves detection sensitivity for minor deviations by amplifying reconstruction errors and uses autoregressive inference for robust multimodal feature learning. By employing a specialized temporal feature extraction module for multi-scale interactive learning, the model efficiently manages redundant data while maintaining high performance. Experimental results on real wind turbine generator (WTG) datasets confirm that the approach significantly outperforms state-of-the-art baselines in both precision and stability.
Key Contributions
- Proposes TransGAN-WT, a hybrid Transformer and GAN model for wind turbine anomaly detection, achieving an average F1 score of 96.10%.
- Introduces a temporal feature extraction module for cross-scale interactive learning, effectively reducing time redundancy in operational data.
- Demonstrates superior reliability over state-of-the-art baselines with a false positive rate (FPR) of 0.06% and statistically validated performance improvements.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2606.03112
- paper_id
- 2606.03112
- paper_source
- openalex
- domain
- time-series
- tags
- anomaly-detectiontime-series-forecastingwind-turbine-monitoringgenerative-adversarial-networkstransformer-models
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-06-05T06:33:12Z