风速取值的准确度对风能的评估有决定性作用,为选择合适的风力机风速插补方法,以内蒙古中部某风电场H为试验风电场,考虑季节及风向影响,划分出风力机轮毂高度风速具有高相关性的风力机分类片区,采用线性回归方法（LR）、随机森林方法（RF）及深度神经网络方法（DNN）,分别对风力机风速进行时空插补及适用性研究。结果表明：风力机插补风速略大于实测风速,LR方法的插补值平均绝对误差、均方根误差分别为0.74、1.00 m/s,RF、DNN方法的插补效果优于LR方法,平均绝对误差减小率分别为9.93%、10.48%,均方根误差减小率分别为8.60%、8.30%。RF、DNN方法在各片区插补效果良好,主导风向片区RF方法最优。按风力机出力情况划分风速,[0, 3）和[12, 25） m/s风速区间推荐使用RF方法,[3, 8）和[8, 12） m/s风速区间更适合采用DNN方法。此外,风力机风速插补误差主要和风速大小及振荡、日变化等有关。
Abstract
The accuracy of wind speed values plays a decisive role in the evaluation of wind energy. In order to select the appropriate wind speed interpolation method for wind turbines, the wind farm H in central Inner Mongolia is used as the test wind farm, and the wind turbine groups, with high correlation of wind speed on the turbine hub height, are divided considering the influence of season and wind direction. The linear regression（LR）, random forests（RF） and deep neural network（DNN） are used to study the spatiotemporal interpolation and applicability of wind speed.The results show that the interpolated wind speed of wind turbine is slightly larger than the measured wind speed. The mean absolute error and root mean square error of interpolation value by LR method are 0.74 and 1.00 m/s, respectively. The interpolation effect of RF and DNN methods is superior to LR method, with average absolute error reduction rates of 9.93% and 10.48%, and root mean square error reduction rates of 8.60% and 8.30%, respectively.The RF and DNN methods have good interpolation effect in each group, and the RF method is optimal in the dominant wind direction group.The wind speed is divided according to the output of the wind turbines. RF method is recommended for the wind speed range of [0, 3） and [12, 25） m/s, and DNN method is more suitable for [3, 8） and [8, 12） m/s. In addition, the interpolation error of wind speed is mainly related to the value, oscillation and daily variation of wind speed.
关键词
风电场 /
风速 /
插补 /
随机森林 /
深度神经网络
Key words
wind farms /
wind speed /
interpolation /
random forests /
deep neural network
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 梁涛, 陈春宇, 谭建鑫, 等. 基于多方面特征提取和迁移学习的风速预测[J]. 太阳能学报, 2023, 44(4): 132-139.
LIANG T, CHEN C Y, TAN J X, et al.Wind speed prediction based on multiple feature extraction and transfer learning[J]. Acta energiae solaris sinica, 2023, 44(4): 132-139.
[2] 杨姝凡. 测风塔的测风数据精度对风电场产能影响的研究[D]. 乌鲁木齐: 新疆大学, 2017.
YANG S F.Research on the influence of wind data accuracy of anemometer tower on wind farm productivity[D]. Urumqi: Xinjiang University, 2017.
[3] 郑侃, 魏煜锋, 文智胜, 等. 基于BP神经网络方法的风电场风速插补分析应用[J]. 南方能源建设, 2021, 8(1): 51-55.
ZHENG K, WEI Y F, WEN Z S, et al.Analysis and application of wind speed interpolation in wind farm based on BP neural n etwork method[J]. Southern energy construction, 2021, 8(1): 51-55.
[4] 秦琼, 刘树洁, 赖旭, 等. GA优化ELM神经网络的风电场测风数据插补[J]. 太阳能学报, 2018, 39(8): 2125-2132.
QIN Q, LIU S J, LAI X, et al.Interpolation of wind speed data in wind farm based on GA optimized elm neural network[J]. Acta energiae solaris sinica, 2018, 39(8): 2125-2132.
[5] 王一博. 风电场运行风速数据清洗与重构方法研究[D]. 北京: 华北电力大学, 2022.
WANG Y B.Research on cleaning and reconstruction method of wind speed data in wind farm operation[D]. Beijing: North China Electric Power University, 2022.
[6] QX/T 645—2022, 风电机组测风资料质量审核与订正[S].
QX/T 645—2022, Quality check and correction of wind observation data from wind turbine[S].
[7] 冯瑞, 杨丽萍, 侯成磊, 等. 基于随机森林的陕西省西安市近地表气温估算[J]. 地球科学与环境学报, 2022, 44(1): 102-113.
FENG R, YANG L P, HOU C L, et al.Estimation of near-surface air temperature in Xi'an city of Shaanxi Province, China based on random forest[J]. Journal of earth sciences and environment, 2022, 44(1): 102-113.
[8] 侯宁, 张晓通, 魏瑜, 等. 基于随机森林方法的中国地表短波辐射估算[J]. 太阳能学报, 2021, 42(2): 31-36.
HOU N, ZHANG X T, WEI Y, et al.Estimation of surface incident shortwave radiation over China based on random forest regression method[J]. Acta energiae solaris sinica, 2021, 42(2): 31-36.
[9] 徐艳平, 陈义安. 基于随机森林回归和气象参数的城市空气质量预测模型: 以重庆市为例[J]. 重庆工商大学学报(自然科学版), 2021, 38(6): 118-124.
XU Y P, CHEN Y A.Urban air quality prediction model based on random forest regression and meteorological parameters: take Chongqing as an example[J]. Journal of Chongqing Technology and Business University (natural science edition), 2021, 38(6): 118-124.
[10] 崔岗, 周广得, 凌骐, 等. 基于POD-DNN代理模型的闸墩锚索有效预应力反演[J]. 水力发电, 2023, 49(9): 53-56, 111.
CUI G, ZHOU G D, LING Q, et al.Inversion of effective prestress of gate pier anchor cables based on POD-DNN surrogate model[J]. Water power, 2023, 49(9): 53-56, 111.
[11] 高宁康, 王小英, 梁嘉烨. 基于随机森林和深度神经网络的恶意域名检测方法[J]. 科学技术创新, 2023(11): 115-118.
GAO N K, WANG X Y, LIANG J Y.Detection of malicious domain names based on random forest and deep neural network[J]. Scientific and technological innovation, 2023(11): 115-118.
[12] GB/T 37523—2019, 风电场气象观测及资料审核、插补与订正技术规范[S].
GB/T 37523—2019, Specification for data inspection and correction of wind power plant meteoroligical observation[S].
[13] 石岚, 徐丽娜, 郝玉珠. 基于风速高相关分区的风电场风速预报订正[J]. 应用气象学报, 2016, 27(4): 506-512.
SHI L, XU L N, HAO Y Z.The correction of forecast wind speed in a wind farm based on partitioning of the high correlation of wind speed[J]. Journal of applied meteorological science, 2016, 27(4): 506-512.
基金
内蒙古自治区自然科学基金项目（2022MS04019）
PDF(1078 KB)
