基于聚类SABO-VMD和组合神经网络的短期光伏发电功率预测  

SHORT-TERM PV POWER FORECASTING BASED ON CLUSTERING SABO-VMD AND ENSEMBLE NEURAL NETWORKS

作　　者：冯建铭 希望·阿不都瓦依提 蔺红[1] Feng Jianming;Xiwang·Abuduwayiti;Lin Hong(School of Electrical Engineering,Xinjiang University,Urumqi 830049,China)

机构地区：[1]新疆大学电气工程学院,乌鲁木齐830049

出　　处：《太阳能学报》2025年第2期357-366,共10页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(52367012)。

摘　　要：针对光伏发电预测单一模型处于不同天气状况时预测精度不高等问题,建立以卷积神经网络(CNN)和双向长短期记忆网络(BiLSTM)为基础的组合神经网络模型。提出一种基于鱼鹰优化算法(OOA),用以优化组合神经网络参数。此外引入注意力机制(Attention)突出强相关性因素的影响。采用高斯混合模型聚类(GMM)划分历史光伏数据为数个天气类型,并提出基于减法平均的优化算法(SABO)优化变分模态分解(VMD)参数,实现对各天气类型数据的分解。实验结果表明:基于SABO-VMD优化数据分解参数能有效提高预测精度;经实验对比分析,该文所提模型精度明显更高。To address the challenges of low prediction accuracy when using a single model for photovoltaic power generation forecasting under varying weather conditions,a composite neural network model is established based on a convolutional neural network and a bidirectional long short-term memory network.The osprey optimization algorithm is introduced to optimize the parameters of ensemble neural networks.Additionally,an attention mechanism is incorporated to emphasize the influence of strong correlation factors.We employ Gaussian mixture model clustering to categorize historical photovoltaic data into various weather types and propose a subtraction average-based optimizer algorithm to optimize variational mode decomposition for data decomposition based on different weather types.Experimental results demonstrate that optimizing data decomposition parameters using SABO-VMD effectively enhances prediction accuracy.In comparison to other combined prediction models,through experimental comparisons,the accuracy of the model proposed in this paper is significantly higher.

关 键 词：光伏功率 变分模态分解 神经网络 功率预测 注意力机制 高斯混合模型聚类 

分 类 号：TM615[电气工程—电力系统及自动化]