| 引用本文: | 孙国强,梁智,俞娜燕,倪晓宇,卫志农,臧海祥,周亦洲.基于EWT和分位数回归森林的短期风电功率概率密度预测[J].电力自动化设备,2018,(8): |
| SUN Guoqiang,LIANG Zhi,YU Nayan,NI Xiaoyu,WEI Zhinong,ZANG Haixiang,ZHOU Yizhou.Short-term wind power probability density forecasting based on EWT and quantile regression forest[J].Electric Power Automation Equipment,2018,(8): |
|
| 摘要: |
| 概率密度预测能够给出未来风电功率可能的波动范围、预测值出现的概率及不确定性等更多信息，提出基于经验小波变换(EWT)和分位数回归森林的短期风电功率概率密度组合预测模型。首先，采用新型自适应信号处理方法——经验小波变换，将原始风电功率序列分解为一系列频率特征互异的经验模式；然后，对每一经验模式序列分别构建分位数回归森林预测模型，得到任意分位点条件下的预测结果，通过叠加不同经验模式预测结果获得最终的短期风电功率预测值；最后，对预测值条件分布采用核密度估计获得任意时刻概率密度预测。仿真结果验证了所提模型的有效性。 |
| 关键词: 经验小波变换 分位数回归森林 核密度估计 概率密度 短期风电功率预测 模型 |
| DOI：10.16081/j.issn.1006-6047.2018.08.023 |
| 分类号:TM761 |
| 基金项目:国家自然科学基金资助项目(51507052)；国家电网公司科技项目(J2017089) |
|
| Short-term wind power probability density forecasting based on EWT and quantile regression forest |
|
SUN Guoqiang1, LIANG Zhi1, YU Nayan2, NI Xiaoyu3, WEI Zhinong1, ZANG Haixiang1, ZHOU Yizhou1
|
|
1.College of Energy and Electrical Engineering, Hohai University, Nanjing 210098, China;2.State Grid Wuxi Power Supply Company, Wuxi 214061, China;3.Wuxi Yang Sheng Technology Co.,Ltd.,Wuxi 214106, China
|
| Abstract: |
| Probability density forecasting can give more information about future wind power output such as possible fluctuation ranges, probability and uncertainty of predicted values and so on. A combined probability density forecasting model for short-term wind power is proposed based on EWT(Empirical Wavelet Transform) and quantile regression forest. Firstly, a new adaptive signal processing method namely empirical wavelet transform is adopted to decompose the original wind power sequence into a series of empirical modes, whose frequency characteristics are different from each other. Then, the quantile regression forest forecasting model is established respectively for each empirical mode, obtaining the predictive results under arbitrary quantiles, and the final wind power forecasting values are obtained by summing up the predictive results of different empirical modes. Finally, the kernel density estimation is applied for the conditional distribution to obtain the probability density forecasting results at any moment. Simulative results verify the effectiveness of the proposed model. |
| Key words: empirical wavelet transform quantile regression forest kernel density estimation probability density short-term wind power forecasting models |
