| 考虑历史相似日与组合权重的光伏功率预测 |
| 点此下载全文 |
| 引用本文：李伟康1,马刚1,袁宇波2,卜强生2,叶志刚2,王伟3,陈遗志3.考虑历史相似日与组合权重的光伏功率预测[J].电网与清洁能源,2025,41(4):87~96 |
| 摘要点击次数: 4 |
| 全文下载次数: 0 |
|
| 基金项目:江苏省重点研发计划项目（BE2022003，BE2022003-5） |
|
| 中文摘要:针对传统预测模型在光伏功率潜在特性的挖掘上存在局限性的问题，提出一种基于相似日聚类及多混合模型组合加权的预测模型。利用Pearson相关性系数选取重要气象特征，并采用模糊C均值聚类将历史日划分为晴天、多云、阴天和雨雪天气4种天气类型；利用完全集合经验模态分解（complete ensemble empirical mode decomposition with adaptive noise，CEEMDAN）将历史光伏功率分解成若干个子序列，并将各个子序列分别通过卷积网络（convolutional neural network，CNN）、长短期记忆网络（long short-term memory，LSTM）与注意力机制（attention）的混合网络模型加以训练；使用灰色关联分析法将（least squares support vector regression，LSSVR）模型的结果与CNN-LSTM-Attention模型的预测结果进行组合，以获得最终的预测结果。算例分析结果表明，所提模型可以提高预测的精度，并能够更好地捕捉天气的波动性。 |
| 中文关键词:光伏功率预测 天气类型 混合模型 灰色关联度 组合权重 |
| |
| Photovoltaic Power Forecasting Considering Historical Similar Days and Ensemble Weighting |
|
|
| Abstract:Addressing the limitations of traditional prediction models in mining the potential characteristics of photovoltaic （PV） power，this paper proposes a prediction model based on similar day clustering and multi-mixed model combination weighting.Firstly，important meteorological features are selected using Pearson correlation coefficients，and fuzzy C-means clustering is used to divide historical days into four weather types： sunny，cloudy，overcast，and rainy/snowy.Secondly，the training sets of the four decomposed weather types are predicted using least squares support vector machine （LSSVR）； furthermore，the historical data is decomposed into several sub-sequences using complete ensemble empirical mode decomposition with adaptive noise（CEEDMAN），and each sub-sequence is predicted using a hybrid network model that incorporates attention mechanisms through convolutional neural networks（CNN） and long short term memory networks （LSTM）； Finally，the grey correlation analysis method is used to combine the prediction results of LSSVR with those of the CNN-LSTM Attention model to obtain the final prediction result.The case analysis shows that the model proposed in this paper can not only improve the prediction accuracy，but also better capture the volatility of weather. |
| keywords:photovoltaic power prediction weather types hybrid model grey correlation degree combination weight |
| 查看全文 查看/发表评论 下载PDF阅读器 |
