针对Stacking算法计算时间较长和样本数据较少的问题，提出了一种基于新向量表示和交叉验证精度加权的改进Stacking算法。采用三层算法结构，第1、2层为初级层，使用随机森林、SVR、XGBoost 3个学习器；第3层为次级层，使用LightGBM对第2层输出再次学习以减弱噪声。用一种新的向量表示法来增大层级之间输入输出数据的样本规模和样本分布密度，来保证数据维度不会随着初级层学习器数目的增多而增大；根据在交叉验证下初级层不同预测模型表现出预测准确度的差异性对结果进行加权处理。利用某光伏电站的发电数据进行实际算例分析，提出的模型在MAE、MSE及$R^2$指标上，相比随机森林和Stacking等模型其预测性能有很大的提升。
Stacking algorithm is good at alleviating over fitting problem in the prediction of photovoltaic power generation, but with drawbacks of long computation time and less sample data. To solve the problem, this paper proposes an improved 3-layer stacking algorithm based on new vector representation and cross validation accuracy weighting. The first and second layers are the primary layer, which use random forest, SVR and XGboost3. The third layer is the secondary layer, and uses LightGBM to learn the output of the second layer again to reduce noise. A new vector representation method is used to increase the sample size and sample distribution density of input and output data between levels to ensure that the data dimension will not increase with the increase of the number of primary level learners. At the same time, the results are weighted according to the difference in the prediction accuracy of different prediction models in the primary layer under cross-validation. Practical analysis is demonstrated by using the power generation data of a photovoltaic power station. Compared with random forest model and Stacking model, the prediction performance of the proposed model has been greatly improved in MAE, MSE and R-Squared.
[1] 赵杰.光伏发电并网系统的相关技术研究[D].天津:天津大学, 2012.
[2] 邓维,刘方明,金海,等.云计算数据中心的新能源应用:研究现状与趋势[J].计算机学报, 2013, 36(3):582-598. Deng W, Liu F M, Jin H, et al. Leveraging renewable energy in cloud computing datacenters:state of the art and future research[J]. Chinese Journal of Computers, 2013, 36(3):582-598.(in Chinese)
[3] 陈倩.分布式光伏并网发电对配电网电能质量的影响[D].上海:上海交通大学, 2015.
[4] 王继东,庞文杰. Fisher判别分类法在光伏并网系统稳态电能质量评估中的应用[J].电力自动化设备, 2017, 37(3):50-54. Wang J D, Pang W J. Application of Fisher discriminant analysis in steady-state power quality evaluation of grid-connected photovoltaic system[J]. Electric Power Automation Equipment, 2017, 37(3):50-54.(in Chinese)
[5] Shi Z C, Liang H, Dinavahi V. Direct interval forecast of uncertain wind power based on recurrent neural networks[J]. IEEE Transactions on Sustainable Energy, 2018, 9(3):1177-1187.
[6] Li Z L, Xia J, Liu A, et al. States prediction for solar power and wind speed using BBA-SVM[J]. IET Renewable Power Generation, 2019, 13(7):1115-1122.
[7] 陈振宇,刘金波,李晨,等.基于LSTM与XGBoost组合模型的超短期电力负荷预测[J].电网技术, 2020, 44(2):614-620. Chen Z Y, Liu J B, Li C, et al. Ultra short-term power load forecasting based on combined LSTM-XGBoost model[J]. Power System Technology, 2020, 44(2):614-620.(in Chinese)
[8] Ju Y, Sun G Y, Chen Q H, et al. A model combining convolutional neural network and LightGBM algorithm for ultra-short-term wind power forecasting[J]. IEEE Access, 2019, 7:28309-28318.
[9] 郑红,叶成,金永红,等.基于Stacking集成学习的流失用户预测方法[J].应用科学学报, 2020, 38(6):944-954. Zheng H, Ye C, Jin Y H, et al. Customer churn prediction method based on stacking ensemble learning[J]. Journal of Applied Sciences, 2020, 38(6):944-954.(in Chinese)
[10] Prasad R, Ali M, Xiang Y, et al. A double decomposition-based modelling approach to forecast weekly solar radiation[J]. Renewable Energy, 2020, 152:9-22.
[11] Banik R, Das P, Ray S, et al. Wind power generation probabilistic modeling using ensemble learning techniques[J]. Materials Today:Proceedings, 2020, 26:2157-2162.
[12] 史佳琪,张建华.基于多模型融合Stacking集成学习方式的负荷预测方法[J].中国电机工程学报, 2019, 39(14):4032-4042. Shi J Q, Zhang J H. Load forecasting based on multi-model by Stacking ensemble learning[J]. Proceedings of the CSEE, 2019, 39(14):4032-4042.(in Chinese)
[13] Wolpert D H. Stacked generalization[J]. Neural Networks, 1992, 5(2):241-259.
[14] 鲁莹,郑少智. Stacking学习与一般集成方法的比较研究[J].中国科技论文在线精品论文, 2018, 11(4):372-379. Lu Y, Zheng S Z. A comparative study of Stacking learning and general ensemble methods[J]. Highlights of Sciencepaper Online, 2018, 11(4):372-379.(in Chinese)
[15] Menahem E, Rokach L, Elovici Y. Troika-an improved stacking schema for classification tasks[J]. Information Sciences, 2009, 179(24):4097-4122.
[16] Abawajy J H, Kelarev A, Chowdhury M. Large iterative multitier ensemble classifiers for security of big data[J]. IEEE Transactions on Emerging Topics in Computing, 2014, 2(3):352-363.
[17] 韦艳艳,李陶深,刘美玲.融合DECORATE的异构分类器集成算法[J].计算机应用研究, 2012, 29(11):4134-4136, 4147. Wei Y Y, Li T S, Liu M L. Algorithm of heterogeneous classifiers ensembles based on DECORATE[J]. Application Research of Computers, 2012, 29(11):4134-4136, 4147.(in Chinese)
[18] Zhou Z H, Jiang Y. NeC4.5:neural ensemble based C4.5[J]. IEEE Transactions on Knowledge and Data Engineering, 2004, 16(6):770-773.
[19] Breiman L. Stacked regressions[J]. Machine Learning, 1996, 24(1):49-64.
[20] 徐慧丽. Stacking算法的研究及改进[D].广州:华南理工大学, 2018.
[21] 徐耀松,段彦强,王雨虹,等.基于相似日选择与改进Stacking集成学习的短期负荷预测[J].传感技术学报, 2020, 33(4):537-545. Xu Y S, Duan Y Q, Wang Y H, et al. Short-term load prediction based on similar day selection and improved Stacking integration learning[J]. Chinese Journal of Sensors and Actuators, 2020, 33(4):537-545.(in Chinese)
