采用最近历史观测值和PLSR进行空间相关性超短期风速预测
发布时间：2018-08-13 11:06
【摘要】：为提高超短期风速预测的可靠性和准确性,将被预测地点(本地)周边测风塔风速风向等当前和最近历史观测值作为基础数据,采用空间相关性来预测本地的未来风速。首先,依据风向和风速的延迟相关性,挑选出上游测风塔。之后,结合最优延迟时间,利用各上游和本地最近的风速观测值来训练预测模型。最后,将各上游风速的当前观测值输入模型,即可得到本地的风速预测值。以偏最小二乘回归(partial least squares regression,PLSR)为主要模型,并采用线性回归(linear regression,LR)、最小二乘支持向量回归等模型进行对照。以冬季风时期的荷兰Huibertgat和天津为被预测地点,进行了PLSR、LR预测误差与模型阶数、样本容量之间关系的数值实验。研究表明,在冬季风时期,当样本容量达到一定程度后,预测误差的变化对阶数、样本容量和模型的类型均不再敏感。这表明空间相关性是一种较为可靠的超短期风速预测方法。
[Abstract]:In order to improve the reliability and accuracy of ultra-short term wind speed prediction, the current and recent historical observations such as wind speed direction of the wind tower around the predicted location (local) are taken as the basic data, and the spatial correlation is used to predict the local future wind speed. First, the upstream wind tower is selected according to the correlation between wind direction and wind speed delay. Then, combined with the optimal delay time, the prediction model is trained by using the most recent observations of the upstream and local wind speeds. Finally, the local wind speed prediction value can be obtained by inputting the current observation values of each upstream wind speed into the model. Partial least square regression (partial least squares regression) was used as the main model, and linear regression and least squares support vector regression (LS-SVM) were used as the control model. A numerical experiment on the relationship between the prediction error and the order of the model and the sample size was carried out using Huibertgat and Tianjin in winter monsoon period as the predicted sites. The results show that the variation of prediction error is not sensitive to order sample size and model type when the sample size reaches a certain level during winter monsoon period. This indicates that spatial correlation is a more reliable method for super-short-term wind speed prediction.
【作者单位】： 天津大学电气与自动化工程学院;天津市过程检测与控制重点实验室(天津大学);
【分类号】：TM614
，
本文编号：2180790
[Abstract]:In order to improve the reliability and accuracy of ultra-short term wind speed prediction, the current and recent historical observations such as wind speed direction of the wind tower around the predicted location (local) are taken as the basic data, and the spatial correlation is used to predict the local future wind speed. First, the upstream wind tower is selected according to the correlation between wind direction and wind speed delay. Then, combined with the optimal delay time, the prediction model is trained by using the most recent observations of the upstream and local wind speeds. Finally, the local wind speed prediction value can be obtained by inputting the current observation values of each upstream wind speed into the model. Partial least square regression (partial least squares regression) was used as the main model, and linear regression and least squares support vector regression (LS-SVM) were used as the control model. A numerical experiment on the relationship between the prediction error and the order of the model and the sample size was carried out using Huibertgat and Tianjin in winter monsoon period as the predicted sites. The results show that the variation of prediction error is not sensitive to order sample size and model type when the sample size reaches a certain level during winter monsoon period. This indicates that spatial correlation is a more reliable method for super-short-term wind speed prediction.
【作者单位】： 天津大学电气与自动化工程学院;天津市过程检测与控制重点实验室(天津大学);
【分类号】：TM614
，
本文编号：2180790
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2180790.html
