基于相似样本和多模型动态最优组合的光伏功率预测 |
点此下载全文 |
引用本文:于红伟1,许国泽2,何旭东3,张锋3,曹晖3.基于相似样本和多模型动态最优组合的光伏功率预测[J].电网与清洁能源,2021,37(9):92~99 |
摘要点击次数: 107 |
全文下载次数: 0 |
|
基金项目:陕西省重点研发计划-国际科技合作计划项目(2019KW-010) |
|
中文摘要:针对传统单一预测方法存在的局限性,引入了考虑特征加权的模糊聚类方法,进行关于天气类型的划分以得到相似样本;提出多模型动态最优组合预测方法,根据各窗口期预测误差的波动情况,设置合适的临近历史样本窗口宽度,利用窗口期中的数据和构建的最优赋权模型进行组合权重的求解,在避免单一预测方法片面性的同时,提高了对各种天气的适应性。通过算例验证分析表明,所提出的组合预测方法在各种天气类型下的预测效果都优于理论预测、BP预测和LSSVM预测等单一预测方法,能够有效提高预测的有效性和准确性,具有较高的工程实用价值。 |
中文关键词:光伏功率 组合预测 相似样本 动态最优 |
|
Photovoltaic Power Prediction Based on Similar Samples and Dynamic Optimal Combination of Multiple Models |
|
|
Abstract:In view of the limitations of the traditional single prediction method, this paper firstly introduces a fuzzy clustering method considering feature weighting to divide the weather type to obtain similar samples; secondly, a multi-model dynamic optimal combination prediction method is proposed, which is based on each window period. For the fluctuation of forecast errors, an appropriate window width for nearby historical samples is set, and the data in the window period and the constructed optimal weighting model are used to solve the combined weights. While avoiding the one-sidedness of a single forecasting method, it also improves the adaptability to various types of weather. Through the verification and analysis of numerical examples, it is proved that the combined forecasting method proposed in this paper is better than the single forecasting methods such as theoretical forecasting, BP forecasting and LSSVM forecasting under various types of weather. It can effectively improve the effectiveness and accuracy of forecasting, which has high engineering practical value. |
keywords:photovoltaic power combined forecasting similar samples dynamic optimization |
查看全文 查看/发表评论 下载PDF阅读器 |