基于多种机器学习算法的风电功率预测融合模型

doi:10.11755/j.issn.1006-7639-2024-05-0710

干旱气象 ›› 2024, Vol. 42 ›› Issue (5): 710-718.DOI: 10.11755/j.issn.1006-7639-2024-05-0710

• “人工智能在干旱气象及相关领域的应用”专栏 • 上一篇下一篇

基于多种机器学习算法的风电功率预测融合模型

韩自奋¹(), 陈宁¹, 范义², 谢志华², 沙莎³()

1.国网甘肃省电力公司，甘肃兰州 730030
2.国网甘肃省电力公司张掖供电公司，甘肃张掖 734000
3.中国气象局兰州干旱气象研究所，甘肃省干旱气候变化与减灾重点实验室，中国气象局干旱气候变化与减灾重点实验室，甘肃兰州 730020

收稿日期:2023-07-21 修回日期:2024-01-29 出版日期:2024-10-31 发布日期:2024-11-17
通讯作者: 沙莎（1985—），女，汉族，辽宁沈阳人，主要从事机器学习应用研究。E-mail：nuist_shasha@126.com。
作者简介:韩自奋（1976—），男，汉族，甘肃民乐人，博士，正高级工程师，主要从事水电及新能源调度管理工作。E-mail: 7638403@qq.com。
基金资助:
国家电网有限公司科技项目“基于合理量率一体下的甘肃风光储最佳配置比例研究”(52270723000900K)

Fusion model of wind power prediction based on multiple machine learning algorithms

HAN Zifen¹(), CHEN Ning¹, FAN Yi², XIE Zhihua², SHA Sha³()

1. State Grid Gansu Electric Power Company， Lanzhou 730030， China
2. State Grid Gansu Electric Power Company Zhangye Power Supply Company， Zhangye 734000， Gansu， China
3. Institute of Arid Meteorology， CMA， Key Laboratory of Arid Climate Change and Reducing Disaster of Gansu Province， Key Laboratory of Arid Climate Change and Disaster Reduction of CMA， Lanzhou 730020， China

Received:2023-07-21 Revised:2024-01-29 Online:2024-10-31 Published:2024-11-17

摘要/Abstract

摘要：

风电场发电功率的精确预测对调度部门及时调整发电规划意义重大。利用2020年1月1日—12月31日甘肃酒泉中广核干东风电场的风功率及风塔风速数据，基于多种典型的机器学习算法和单一模型在测试集上表现出的特点，研究K近邻（K-Nearest Neighbor，KNN）、装袋算法（Bootstrap Aggregating，BA）与卷积神经网络（Convolutional Neural Network，CNN）的组合方法，建立融合KNN、BA与CNN的风电功率预测模型。结果表明，单一模型均存在对部分值低值高估的情况，其中多层感知机（Multilayer Perceptron，MLP）、CNN这两种神经网络还存在明显的高值低估现象；BA模型预测精度最高，其在测试集上的均方根误差（Root Mean Square Error， RMSE）为13.08 MW；组合模型均能一定程度上提高单一模型的预测精度，其中CNN组合模型在测试集上RMSE为12.21 MW，比单一模型中最好的BA模型RMSE下降约6.7%，CNN组合模型可以明显改善CNN单一模型高值低估、低值高估和BA模型低值高估的情况。

关键词: CNN, 装袋算法, KNN, 风电功率

Abstract:

The accurate prediction of wind power is of great significance for the dispatching department to adjust power generation planning in a timely manner. Machine learning is one of the main methods of wind power prediction at present. However, how to select reliable and effective single model from numerous machine learning algorithms, and how to fit and associate different models are the key and difficult points of multi-model combination prediction methods. Based on the wind power and wind tower wind speed data of Jiuquan Gandong Wind Power Electric Station of China Genneral Nuclear Power Corporation from January 1, 2020 to December 31, 2020, and the characteristics of various typical machine learning algorithms and a single model on the test set, the combination method of K-Nearest Neighbor (KNN), Bootstrap Aggregating (BA) and Convolutional Neural Network (CNN) is studied, and a wind power prediction model that integrates KNN, BA and CNN is established. The results show that all single models overestimate the low values of some values, and the Multi-Layer Perceptron (MLP) and CNN neural networks also underestimate the high values. The BA model has the highest prediction accuracy, and its Root Mean Square Error (RMSE) on the test set is 13.08 MW. The combined models can improve the prediction accuracy of the single model to a certain extent. The RMSE of the CNN combined model on the test set is 12.21 MW, which is about 6.7% lower than the RMSE of the best BA model in the single models. The CNN combined model can significantly improve the situation that the high value is underestimated, the low value is overestimated for the CNN single model, and the low value is overestimated for the BA model. The prediction model established in this article can be extended to practical wind power prediction.

Key words: CNN, Bagging, KNN, wind power

中图分类号:

TM614

韩自奋, 陈宁, 范义, 谢志华, 沙莎. 基于多种机器学习算法的风电功率预测融合模型[J]. 干旱气象, 2024, 42(5): 710-718.

HAN Zifen, CHEN Ning, FAN Yi, XIE Zhihua, SHA Sha. Fusion model of wind power prediction based on multiple machine learning algorithms[J]. Journal of Arid Meteorology, 2024, 42(5): 710-718.

图/表 11

图1 多层感知机示意图

Fig.1 The diagram of MLP

图2 卷积神经网络示意图

Fig.2 The diagram of CNN

图3 各变量之间的相关性热力图

Fig.3 The correlation heatmap between various variables

图4 不同高度层风速之间的关系（a）10 m与30 m，（b）30 m与50 m，（c）50 m与70 m

Fig.4 The relationship between wind speeds at different levels （a） 10 m and 30 m，（b） 30 m and 50 m，（c） 50 m and 70 m

图5 70 m高度层风向（a）及风速（b）与风功率的关系

Fig.5 Relationship between wind direction （a）， wind speed （b） and wind power at the height of 70 m

表1 各种机器学习算法模型在训练集和测试集上的拟合效果

Tab.1 The fitting results of various machine learning algorithm models on the training and testing set

模型	训练集		测试集
模型	RMSE/MW	R²	RMSE/MW	R²
KNN	12.92	0.93	17.21	0.88
GBDT	13.39	0.93	15.32	0.90
AB	17.54	0.87	18.57	0.86
BA	4.73	0.99	13.08	0.93
MLP	16.33	0.89	17.00	0.88
CNN	14.37	0.91	15.19	0.90

图6 各模型在测试集上实际风功率和预测风功率的散点图

Fig.6 Scatter plots of actual wind power and predicted wind power for each model on test data

图7 BA、GDBT、CNN模型在部分测试集数据上的折线图

Fig.7 Line charts of BA， GDBT， CNN models on partial test set data

表2 组合模型在训练集和测试集上的拟合效果

Tab.2 The fitting results of the combined models on the training and testing sets

模型	训练集		测试集
模型	RMSE/MW	R²	RMSE/MW	R²
KNN+BA	3.73	0.99	12.79	0.93
BA+KNN	3.32	0.99	13.10	0.93
CNN+BA	4.38	0.99	12.91	0.93
CNN组合模型	3.68	0.99	12.21	0.94

图8 组合模型在测试集数据上的拟合散点图

Fig.8 Scatter plots of actual wind power and predicted wind power for combined models on test data

图9 KNN+BA、BA+KNN、CNN+BA、CNN组合模型在部分测试集数据上的折线图

Fig.9 Line charts of KNN+BA， BA+KNN， CNN+BA， CNN combined model on partial test set data

参考文献 30

[1]	蔡义勇, 刘爱鸣, 陈雪钦, 等, 2006. 福建省热带气旋暴雨落区完全预报方法[J]. 气象科技, 34(2): 132-137.
[2]	丁家乐, 陈国初, 原阔, 2019. 基于改进萤火虫算法的短期风功率预测[J]. 系统仿真学报, 31(11): 2 509-2 516.
[3]	谷学静, 陈洪磊, 孙泽贤, 等, 2023. 基于VMD-RL-LSTM的短期风功率预测[J]. 计算机仿真, 40(4): 89-93.
[4]	贾睿, 杨国华, 郑豪丰, 等, 2022. 基于自适应权重的CNN-LSTM &GRU组合风电功率预测方法[J]. 中国电力, 55(5): 47-56.
[5]	李国, 江晓东, 2018. 基于提升回归树与随机森林的风电功率集成预测方法[J]. 电力系统及其自动化学报, 30(11): 70-74.
[6]	梁涛, 石欢, 崔洁, 等, 2020. 基于Bagging神经网络集成的风功率预测[J]. 水电能源科学, 38(4): 205-208.
[7]	林卫星, 文劲宇, 艾小猛, 等, 2012. 风电功率波动特性的概率分布研究[J]. 中国电机工程学报, 32(1): 38-46.
[8]	牟银杰, 2003. 完全预报法在高原降水预报中的应用[J]. 四川气象, 23(2): 16-18.
[9]	牛哲文, 余泽远, 李波, 等, 2018. 基于深度门控循环单元神经网络的短期风功率预测模型[J]. 电力自动化设备, 38(5): 36-42.
[10]	申艳杰, 尚兆晖, 2023. 基于深度学习的多尺度风功率预测的研究[J]. 自动化应用, 64(7): 87-90.
[11]	宋江涛, 崔双喜, 刘洪广, 2023. 基于二次分解NGO-VMD残差项与长短时记忆神经网络的超短期风功率预测[J]. 科学技术与工程, 23(6): 2 428-2 437.
[12]	王雨城, 2020. 基于深度神经网络的风功率预测模型[D]. 上海: 上海电机学院.
[13]	许沛华, 陈正洪, 孙延维, 等, 2021. 湖北山区复杂地形条件下风电功率预报算法研究[J]. 干旱气象, 39(3): 524-532.
[14]	杨秀媛, 裘微江, 金鑫城, 等, 2018. 改进K近邻算法在风功率预测及风水协同运行中的应用[J]. 电网技术, 42(3): 772-778.
[15]	余贻鑫, 秦超, 2014. 智能电网基本理念阐释[J]. 中国科学:信息科学, 44(6): 694-701.
[16]	张铁军, 朱蓉, 李照荣, 等, 2011. 利用中尺度模式与诊断风场模型进行风电功率预报[J]. 干旱气象, 29(4): 500-503.
[17]	赵征, 汪向硕, 2020. 基于CEEMD和改进时间序列模型的超短期风功率多步预测[J]. 太阳能学报, 41(7): 352-358.
[18]	邹金, 朱继忠, 赖旭, 等, 2019. 基于时空自回归移动平均模型的风电出力序列模拟[J]. 电力系统自动化, 43(3): 101-107.
[19]	BREIMAN L, 1996. Bagging predictors[J]. Machine Learning, 24(2):123-140. DOI:10.1007/BF00058655.
[20]	CAO Y K, HU Q Q, SHI H, et al, 2019. Prediction of wind power generation base on neural network in consideration of the fault time[J]. IEEJ Transactions on Electrical and Electronic Engineering, 14(5): 670-679. DOI: 10.1002/tee.22853.
[21]	COVER T, HART P, 1967. Nearest neighbor pattern classification[J]. IEEE Transactions on Information Theory, 13(1):21-27. DOI:10.1109/TIT.1967.1053964.
[22]	ERDEM E, SHI J, 2011. ARMA based approaches for forecasting the tuple of wind speed and direction[J]. Applied Energy, 88(4): 1 405-1 414. DOI: 10.1016/j.apenergy.2010.10.031.
[23]	FREUND Y, SCHAPIRE R E, 1997. A decision-theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences, 55(1):119-139. DOI: 10.1006/jcss.1997.1504.
[24]	FRIEDMAN J H, 2001. Greedy function approximation: A gradient boosting machine[J]. The Annals of Statistics, 29(5): 1 189-1 232. DOI:10.1214/aos/1013203451.
[25]	HONG Y Y, RIOFLORIDO C L P P, 2019. A hybrid deep learning-based neural network for 24-h ahead wind power forecasting[J]. Applied Energy, 250: 530-539. DOI: 10.1016/j.apenergy.2019.05.044.
[26]	LECUN Y, BOTTOU L, BENGIO Y, et al, 1998. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 86(11): 2 278-2 324. DOI:10.1109/5.726791.
[27]	LIU H, MI X W, LI Y F, 2018. Smart deep learning based wind speed prediction model using wavelet packet decomposition, convolutional neural network and convolutional long short term memory network[J]. Energy Conversion and Management, 166: 120-131. DOI:10.1016/j.enconman.2018.04.021.
[28]	NIU Z W, YU Z Y, TANG W H, et al, 2020. Wind power forecasting using attention-based gated recurrent unit network[J]. Energy, 196: 117081. DOI: 10.1016/j.energy.2020.117081.
[29]	RUMELHART D E, HINTON G E, WILLIAMS R J, 1986. Learning representations by back propagating errors[J]. Nature, 323(6088):533-536. DOI:10.1038/323533a0.
[30]	WANG K J, QI X X, LIU H D, et al, 2018. Deep belief network based k-means cluster approach for short-term wind power forecasting[J]. Energy, 165: 840-852. DOI:10.1016/j.energy.2018.09.118.

基于多种机器学习算法的风电功率预测融合模型

Fusion model of wind power prediction based on multiple machine learning algorithms

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 30

相关文章 1

编辑推荐

Metrics