文章摘要
叶林,路朋,滕景竹,等.考虑风电功率爬坡的功率预测-校正模型[J].电力系统自动化,2019,43(6):49-56. DOI: 10.7500/AEPS20180321004.
YE Lin,LU Peng,TENG Jingzhu, et al.Power Prediction and Correction Model Considering Wind Power Ramping Events[J].Automation of Electric Power Systems,2019,43(6):49-56. DOI: 10.7500/AEPS20180321004.
考虑风电功率爬坡的功率预测-校正模型
Power Prediction and Correction Model Considering Wind Power Ramping Events
DOI:10.7500/AEPS20180321004
关键词: 风电功率爬坡事件  最优旋转门算法  极限学习机  风电功率预测
KeyWords: wind power ramping event(WPRE)  optimized swinging door algorithm  extreme learning machine(ELM)  wind power forecasting
上网日期:2019-01-31
基金项目:国家电网公司科技项目(5201011600TS);国家自然科学基金资助项目(51677188);国家自然科学基金中英国际合作交流基金资助项目(51711530227)
作者单位E-mail
叶林 中国农业大学信息与电气工程学院, 北京市 100083 yelin@cau.edu.cn 
路朋 中国农业大学信息与电气工程学院, 北京市 100083  
滕景竹 中国农业大学信息与电气工程学院, 北京市 100083  
翟丙旭 国网冀北电力有限公司电力调度控制中心, 北京市 100055  
吴林林 国网冀北电力有限公司电力科学研究院, 北京市 100045  
蓝海波 国网冀北电力有限公司电力调度控制中心, 北京市 100055  
仲悟之 中国电力科学研究院有限公司, 北京市 100192  
摘要:
      随着大规模风电接入电力系统,风电功率爬坡事件对电网的安全稳定运行带来一定的影响。研究爬坡事件发生时的功率预测已越来越迫切。基于极限学习机理论,提出了一种考虑风电功率爬坡事件的超短期功率预测和校正模型。首先,利用最优旋转门算法对当前爬坡事件进行识别,提取爬坡事件特征值,建立模糊C均值聚类模型以得到同类数据,在此基础上,采用极限学习机算法对上述数据进行训练、预测,通过元组向量时间扭曲法在历史风电功率预测爬坡事件库中寻找与当前风电功率预测结果相似的爬坡事件,得到功率预测历史相似爬坡事件。最后,利用功率预测历史匹配值与实际值之间的特征值误差,对风电功率预测结果进行修正。算例表明,所提方法可准确识别风电功率爬坡事件、有效提高风电功率超短期预测精度。
Abstract:
      With the integration of large-scale wind power into power system, wind power ramping events(WPREs)have brought series of effect on the stability and safety for power system operation. It is becoming more and more urgent to study the power prediction when the ramping event occurs. Ultra-short-term prediction and correction model considering the WPREs is employed based on extreme learning machine(ELM). Firstly, the optimized swinging door algorithm is utilized to identify the WPREs and extract the feature value of WPREs. Then, the fuzzy C-means clustering model is established to obtain the congeneric data. On this basis, by using ELM to train and predict the data mentioned above, similar ramping events to the results of current wind power prediction are found from historical WPREs library by tuple vector time warp(TVTW)algorithm, which obtains historically similar ramping events of power prediction. Finally, the wind power prediction result is corrected corresponding to the difference between the historical value and the actual value of predicted power. Case studies show that the proposed method can accurately identify WPREs and significantly improve the precision of ultra-short-term forecasting for wind power.
查看全文(Free!)   查看附录   查看/发表评论  下载PDF阅读器