摘要
利用陕西省某一风电场区域内的观测资料,分析了该风电场的风速规律,并引入最优训练期方案,研究利用线性回归方法建立风电功率预报模型的可行性。结果表明:该风电场区域,不同高度的风速及其高度间的风速差异均表现出最大值出现在夜间,最小值出现在白天,从低层到高层的风速日变化趋势一致的特征。一日中,风速与风电功率在09:00~17:00时段的相关系数明显小于其它时段。按照风速是否大于5 m·s^-1将训练期观测样本分为2组,可以明显改善风速与风电功率的回归关系。以风机轮毂高度处的风速作为预报因子,并引入风电功率与风速之间相关系数的日变化规律、以及不同风速量级下风速与风电功率之间回归关系的差异性,采用最优训练期方案和一元线性回归方法建立的风电功率预报方程,具有预报误差小和最优训练期短的特点,满足实际业务需求。
Based on the observation data of one wind tower and thirty three wind turbines of a wind farm in Shaan’xi Province from 1st January 2016 to 31st December 2018,wind speed variation in the wind farm was analyzed,forecast models for wind power business were established by using simple and multiple linear regression methods with introduc⁃ing the optimal training period.The optimal training period was defined as a sliding cycle,which was the N days before the forecast date in the training samples,for n=4,5…,365 days,the N was decided by the minimum absolute error of the forecast,and then the wind power forecast model was established with the optimal training period day by day.The re⁃sults showed that the wind speed and its difference at various heights had a clear diurnal cycle characteristics from the land surface to height of 80 meters,which exhibited that the maximum value took place in the night time,while the mini⁃mum value took place in the day time.The wind speed difference between heights of 80 meters and 10 meters always maintained the maximum in all levels,then the second maximum was between heights of 50 meters and 10 meters,but the wind speed difference was not completely proportional with the height difference between them.For example,the vari⁃ation in wind speed between height of 30 meters and 10 meters was bigger than that between 30 meters and 80 meters.Wind speed was the most important factor affecting wind power,and wind power had a significant positive correlation with wind speed at various heights.The correlation between wind speed and wind power during the 09:00 and 17:00 in every day was significantly better than the other period of time.If the observation samples were divided into two groups according to whether the wind speed was greater than 5 m·s-1,then regression relationship between wind speed and wind power would be better,and the wind power forecast model would be significantly improved.Based on diurnal variation of correlation coefficient between wind power output and wind s
作者
王丹
高红燕
杨艳超
李博
张黎
WANG Dan;GAO Hong-yan;YANG Yan-chao;LI Bo;ZHANG Li(Shaanxi Meteological Service Center,Xi’an 710014,Shaanxi,China;Guohua Energy Investment Co.,Ltd.(Shaanxi Branch),Yulin 719000,Shaanxi,China;Shaanxi Meteological Observatory,Xi’an 710014,Shaanxi,China)
出处
《干旱区地理》
CSCD
北大核心
2020年第5期1261-1269,共9页
Arid Land Geography
基金
陕西省自然科学基金项目(2019JM-342)
中国气象局预报员专项项目(CMAYBY2019-117)
陕西省气象服务中心“电力气象服务技术研发创新团队”共同资助。
关键词
风电场
风电功率预报
线性回归方法
最优训练期
wind farm
wind power prediction
linear regression method
optimal training period
