基于VMD-TSAO-BiLSTM的短期光伏发电功率预测  

Short-term PV power prediction based on VMD-TSAO-BiLSTM

作　　者：李有为 王芳(指导)[1] 顾伟光 LI Youwei;WANG Fang;GU Weiguang(School of Electrical Engineering,Shanghai Dianji University,Shanghai 201306,China)

机构地区：[1]上海电机学院电气学院,上海201306

出　　处：《上海电机学院学报》2025年第1期21-26,共6页Journal of Shanghai Dianji University

摘　　要：光伏发电对于解决全球能源短缺问题具有重要意义,准确预测光伏发电功率有助于光电并网的合理调度和可靠的电网运行。提出了一种基于变分模态分解(VMD)、改进的雪消融优化算法(TSAO)以及双向长短期记忆神经网络(BiLSTM)的短期光伏发电功率预测模型。首先,原始光伏功率经VMD分解为多个子模态;然后,对每个子模态分别建立TSAO-BiLSTM预测模型,使用改进的Tent混沌映射、自适应t分布和动态选择策略对雪消融优化算法(SAO)进行改进,使用改进后的SAO对BiLSTM的初始学习率、最大训练周期、隐藏单元数目以及L2正则化参数进行寻优;最后,将各个子模型的预测结果叠加得到最终预测结果。仿真结果表明:VMD-TSAO-BiLSTM模型与其他模型相比,能更好地拟合光伏功率数据,具有较高的预测精度。Photovoltaic(PV)power generation is significant in addressing the global energy shortage.Accurate prediction of PV power generation can aid in efficient scheduling photovoltaic grid-connection and ensuring reliable grid operation.This paper proposes a short-term PV power prediction model based on variational modal decomposition(VMD),the improved snow ablation optimization(TSAO)algorithm,and the bidirectional long short-term memory(BiLSTM)network.The raw PV power is first decomposed into multiple sub-modalities using VMD.Then,a TSAOBiLSTM prediction model is established for each sub-modality separately.The snow ablation optimisation(SAO)algorithm is improved by using the enhanced tent chaos mapping,adaptive t-distribution and a dynamic selection strategy.The improved SAO is used to optimize the BiLSTM model by tuning its initial learning rate,maximum training epochs,number of hidden units,and L2 regularization parameters.The results of each sub-model are then combined to obtain the final prediction results.The simulation results demonstrate that the VMD-TSAO-BiLSTM model outperforms other models in terms of fitting the PV power data, and exhibits high predictionaccuracy.

关 键 词：光伏功率预测 变分模态分解 优化算法 双向长短期记忆网络 

分 类 号：TM615[电气工程—电力系统及自动化]