摘要
在“双碳”背景下,新能源发电功率的准确预测对于电力系统的平稳运行至关重要。提出了一种自适应性的VMD-Stacking集成模型,以解决数据集变化时传统学习模型预测精度不高的问题。利用皮尔逊相关系数选择与发电功率强相关的气象特征,通过变分模态分解(Variational mode decomposition,VMD)将功率数据分解为多个模态分量,由此构成新的数据集。运用贝叶斯优化算法调整超参数,综合评判随机森林等8种学习模型的评价指标,自适应选出预测性能最优的3种模型作为基学习器,并选用稳定性和泛化能力相对较强的线性回归(Linear Regression)作为元学习器,建立Stacking融合模型。对各分量的预测值叠加,得到最终预测结果。以某新能源场站为例,对风、光电站的发电功率进行预测。算例验证结果表明,该模型在面对不同数据集时,体现出较强的适应性,预测性能也得到显著的提升。
In the context of“dual carbon”,accurate prediction of new energy generation power is essential for ensuring the smooth operation of the power system.An adaptive VMD-Stacking integrated model is proposed to address the problem of low prediction accuracy of traditional learning models when the dataset changes.Using Pearson correlation coefficient to select meteorological features strongly correlated with power generation,and variational mode decomposition is used to decompose power datas into multiple modal components,creating a new dataset.The Bayes optimization algorithm is employed to adjust hyperparameters and comprehensively evaluate evaluation indexes for eight learning models including random forest.Three models demonstrating superior predictive performance are adaptively selected as base learners,and use linear regression,which has relatively strong stability and generalization ability,as the meta-learner to establish the Stacking fusion model.The predicted value for each component is aggregated to obtain the final predicted result.Using a new energy station as an example,the power generation of wind and solar station are predicted.Simulation results demonstrate that this model exhibits strong adaptability across different datasets and significantly improves prediction performance.
作者
慈铁军
廖子恒
任梦晨
梁音
吴自高
CI Tiejun;LIAO Ziheng;REN Mengchen;LIANG Yin;WU Zigao(College of Energy Power and Mechanical Engineering,North China Electric Power University,Baoding 071003,China;Hebei Key Laboratory of Electric Machinery Health Maintenance&Failure Prevention,North China Electric Power university,Baoding 071003,China)
出处
《电力科学与工程》
2024年第9期14-23,共10页
Electric Power Science and Engineering
基金
国家自然科学基金资助项目(52275109)
河北省教育厅在读研究生创新能力培养资助项目(CXZZSS2024164)。
