基于聚类的HPO-BILSTM光伏功率短期预测  

CLUSTERING-BASED HPO-BILSTM SHORT-TERM PREDICTION OF PV POWER

作　　者：周育才[1] 肖添 谢七月 付强 钟敏 Zhou Yucai;Xiao Tian;Xie Qiyue;Fu Qiang;Zhong Min(School of Electrical and Information Engineering,Changsha University of Science&Technology,Changsha 410114,China;School of Energy and Power Engineering,Changsha University of Science&Technology,Changsha 410114,China)

机构地区：[1]长沙理工大学电气与信息工程学院,长沙410114 [2]长沙理工大学能源与动力工程学院,长沙410114

出　　处：《太阳能学报》2024年第4期512-518,共7页Acta Energiae Solaris Sinica

摘　　要：考虑到光伏发电功率在不同天气类型下的波动性和不确定性,对此提出一种基于模糊C均值聚类算法(FCM)和猎食者优化算法(HPO)优化双向长短期记忆网络(BILSTM)的光伏发电短期功率预测模型。首先对光伏发电数据进行处理和分析,再进行主成分分析(PCA)降维和FCM聚类算法将数据按天气类型分为阴、晴、雨;最后通过HPO筛选得出BILSTM神经网络的最佳超参数,避免因超参数设置不佳对实验带来的影响,进一步提高实验的准确性和模型的泛化能力。最后通过预测和对比实验进行分析,验证所提方法的优越性。Considering the volatility and uncertainty of PV power generation under different weather types,a short-term power prediction model of PV power generation based on fuzzy C-mean clustering algorithm and predator optimization algorithm to optimize the bi-directional long short-term memory network is proposed.Firstly,the PV power generation data are processed and analyzed,then the principal component analysis downscaling and FCM clustering algorithm are performed to classify the data into cloudy,sunny and rainy according to weather types.Then,the best hyperparameters of the BILSTM neural network are derived through HPO screening,which avoids the impact of poor hyperparameter settings on the experiments and further improves the accuracy of the experiments and the generalization ability of the model.Finally,the superiority of the proposed method is verified by prediction and comparison experiments.

关 键 词：光伏发电 双向长短期记忆网络 功率预测 降维 聚类 优化算法 

分 类 号：TM615[电气工程—电力系统及自动化]