| 引用本文: | 梁志峰,朱立轩,李彪,王清凉,陈大玮.台风天气下风电功率预测偏差超短期预警与误差修正方法[J].电力自动化设备,2025,45(12):66-73. |
| LIANG Zhifeng,ZHU Lixuan,LI Biao,WANG Qingliang,CHEN Dawei.Ultra-short-term warning and error correction method for wind power forecasting deviation during typhoon[J].Electric Power Automation Equipment,2025,45(12):66-73. |
|
| 摘要: |
| 针对台风天气下复杂气象条件导致风电功率预测偏差增大、精度下降的问题，提出了一种考虑气象功率耦合特性及误差修正的预测偏差预警方法。挖掘典型台风场景，基于时频域多维特征揭示台风天气对风电功率的总体影响机制。建立基于灰狼优化的极限学习机(GWO-ELM)模型的风电功率预测偏差超短期预警模型，在保证计算效率的同时提升稳定性，实现针对台风天气下风电功率突变、爬坡等事件的可靠评估和预警。结合预测模型的误差特性，采用改进的精度提升模型降低预测时延的影响，改善风电功率预测偏差预警效果。选取我国南部某省风电功率数据进行对比实验，验证所提方法的有效性。实验结果表明，所提方法可实现台风天气下风电功率预测偏差的准确、高效预警，融合误差特性的改进模型可进一步提升预测偏差评估精度。在4种不同的台风场景下，所提方法的平均预测准确率均高于96.4 %，与常规方法相比可提升约1%且计算成本更低。 |
| 关键词: 台风天气 风电功率预测 预测偏差预警 超短期预测 误差修正 机器学习 |
| DOI：10.16081/j.epae.202509021 |
| 分类号:TM73 |
| 基金项目: |
|
| Ultra-short-term warning and error correction method for wind power forecasting deviation during typhoon |
|
LIANG Zhifeng1,2, ZHU Lixuan3, LI Biao3, WANG Qingliang4, CHEN Dawei3
|
|
1.Department of Electrical Engineering, Tsinghua University, Beijing 100084, China;2.National Electric Power Dispatching and Control Center of SGCC, Beijing 100031, China;3.State Grid Fujian Electric Power Research Institute, Fuzhou 350003, China;4.State Grid Fujian Power Dispatching Control Center, Fuzhou 350003, China
|
| Abstract: |
| Complex meteorological conditions during typhoon weather often lead to increased error and decreased accuracy in wind power forecasting. To address this problem, a forecasting deviation warning method considering weather-power coupling characteristics and error correction is proposed. By mining typical typhoon scenarios, the overall impact mechanism of typhoon weather on wind power is revealed based on multi-dimensional features in time-frequency domain. A ultra-short-term warning model for wind power forecasting deviation is established based on extreme learning machine based on grey wolf optimizer(GWO-ELM) model, which enhances stability while maintaining computational efficiency, enabling reliable assessment and warning for events such as abrupt power changes and ramping during typhoon weather. By incorporating the error characteristics of the forecasting model, an improved accuracy-enhancement model is employed to mitigate the impact of prediction latency and enhance the performance of wind power forecasting deviation warning. Comparative experiments are conducted using provincial wind power data from a province in Southern China to validate the effectiveness of the proposed method. The experimental results demonstrate that the proposed method achieves accurate and efficient warning of wind power forecasting deviation during typhoon weather. Moreover, the improved model integrating error characteristics further enhances the accuracy of deviation assessment. Under four different typhoon scenarios, the proposed method achieves an average prediction accuracy exceeding 96.4 %,and represents an improvement of approximately 1% with lower computational costs compared with the conventional methods. |
| Key words: typhoon weather wind power prediction forecasting deviation warning ultra-short-term forecasting error correction machine learning |
