基于时序迁移学习算法的有源配电网多区域分布式光伏功率预测方法

Multi area Distributed Photovoltaic Power Prediction Method for Active Distribution Networks Based on Temporal Transfer Learning Algorithm

下载PDF

导出

摘要随着分布式光伏发电在有源配电网中的规模化应用,对光伏功率进行准确预测成为提高电网运行效率和可靠性的关键问题。然而,由于不同区域的光照条件、天气变化等因素的差异,传统的功率预测方法在多区域分布式光伏发电中存在一定局限性。为解决这一问题,文章提出了一种基于时序迁移学习算法的有源配电网多区域分布式光伏功率预测方法。首先,文章对分布式光伏时序功率进行预测采用了长短期记忆网络(Long Short Term Memory Network,LSTM)算法,该算法能够有效捕捉时序数据的长期依赖关系,适用于光伏功率的时序预测。其次,针对不同区域光伏发电数据的差异性,本文引入迁移学习的思想,将训练数据较多的区域的分布式光伏模型迁移至训练数据较少的区域,以提高预测模型的泛化能力和准确性。文章利用算例仿真分析验证了时序迁移学习算法在预测有源配电网多区域分布式光伏功率的有效性和实用性。 With the large-scale application of distributed photovoltaic power generation in active distribution networks,accurate prediction of photovoltaic power has become a key issue in improving the efficiency and reliability of power grid operation.However,due to differences in lighting conditions,weather changes,and other factors in different regions,traditional power prediction methods have certain limitations in multi regional distributed photovoltaic power generation.To address this ssue,this paper proposes a multi area distributed photovoltaic power prediction method for active distribution networks based on temporal transfer learning algorithm.Firstly,this article adopts the Long Short Term Memory Network(LSTM)algorithm to predict the distributed photovoltaic temporal power,which can effectively capture the long-term dependency relationship of temporal data and is suitable for temporal prediction of photovoltaic power.Secondly,in response to the differences in photovoltaic power generation data in differcnt regions,this article introduccs the idea of transfer learning to transfer distributed photovoltaic models from regions with more training data to regions with less training data,in order to improve the generalization ability and accuracy of the prediction model.This article uses numerical simulation analysis to verify the effectiveness and practicality of the temporal transfer learning algorithm in predicting distributed photovoltaic power in multiple arcas of active distribution networks.

作者刘博文 LIU Bowen(State Grid Jiangsu Electric Power Co.,Ltd.Guanyun County Power Supply Branch,Jiangsu Guanyun 222200)

机构地区国网江苏省电力有限公司灌云县供电分公司

出处《长江信息通信》 2024年第8期119-121,共3页 Changjiang Information & Communications

关键词有源配电网迁移学习分布式光伏功率预测长短期记忆网络象群迁徙算法 Active distribution network Transfer learning Distributed photovoltaics Power prediction Long short-term memory network Elephant herding optimization

分类号 TM734 [电气工程—电力系统及自动化]