Wind Forecast Using Extension of Conditional Generative Adversarial Networks for Optimal Wind Turbine Yaw Control Systems
碩士 === 國立交通大學 === 數據科學與工程研究所 === 107 === The development of offshore wind power generation has been progressing at a rapid pace in recent years. In order to capture wind energy more efficiently and ensure lower maintenance costs of the wind turbines, accurate wind forecasts are needed for the optimi...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2019
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/w6tg9u |
| Summary: | 碩士 === 國立交通大學 === 數據科學與工程研究所 === 107 === The development of offshore wind power generation has been progressing at a rapid pace in recent years. In order to capture wind energy more efficiently and ensure lower maintenance costs of the wind turbines, accurate wind forecasts are needed for the optimization of wind turbine yaw control systems. A convolutional LSTM is a typical deep learning generative model for spatiotemporal sequence predictions. However, it does not achieve good performance for the complex dataset in the study. We design the deep learning model based on conditional GANs with the optimized loss function to provide more accurate spatiotemporal sequences of wind forecast results. Furthermore, we are developing the reinforcement-learning-based method to optimize the yaw control system to maximize wind energy capture. Preliminary results have shown that our forecasting approach is useful to improve the scheduling design of wind turbine optimal yaw control systems.
|
|---|