摘要
随着电动船舶的发展和普及,内河流域的港口用能结构正逐步由燃油转变为清洁的电能,港口负荷将对配电网峰谷差造成显著影响。为准确描述电动船舶充电负荷特征,提出了一种基于改进鲁棒性辅助分类生成对抗网络(RAC-GAN)的电动船舶充电负荷场景生成方法。首先,分析电动船舶充电负荷的特征,构建含环境特征与充电负荷的原始数据集;然后,对RAC-GAN进行改进,加入变分编码器对船舶数据集进行降维,抽取特征信息簇标签,并在判别器中引入噪声过渡模型和卷积层,以提高判别器的抗噪能力,并对网络的损失函数进行重定义;最后,以中国实际港口为例,基于改进的RAC-GAN生成船舶充电负荷的海量场景。仿真结果表明,所提方法能够学习到电动船舶的负荷特征,对噪声具有较高的鲁棒性,并且可以有效生成大量满足真实样本概率分布特征的电动船舶充电负荷场景。
With the development and popularization of electric ships,the energy structure of ports in inland river basins is gradually changing from fuel oil to clean electric energy,and the port loads will significantly impact the peak-to-valley difference of the distribution network.To accurately characterize the charging loads of electric ships,a scenario generation method for electric ship charging loads based on the improved robust auxiliary classifier generation adversarial network(RAC-GAN)is proposed.First,the characteristics of electric ship charging loads are analyzed,and the original dataset containing environmental characteristics and charging loads is constructed.Then,the RAC-GAN is improved by adding a variational encoder to reduce the dimensionality of the ship dataset,extracting the feature information cluster labels,and introducing a noise transition model and a convolutional layer to improve the resistance to noise for the discriminator,and the loss function of the network is redefined.Finally,taking an actual port in China as a case,massive scenarios of ship charging loads are generated based on the improved RAC-GAN.Simulation results show that the proposed method can learn the load characteristics of electric ships,has high robustness to noise,and can effectively develop many electric ship charging load scenarios that satisfy the probability distribution characteristics of actual samples.
作者
廖菲
杨军
林毅
薛静玮
吴少将
朱睿
LIAO Fei;YANG Jun;LIN Yi;XUE Jingwei;WU Shaojiang;ZHU Rui(Hubei Engineering and Technology Research Center for AC/DC Intelligent Distribution Network,Wuhan 430072,China;School of Electrical Engineering and Automation,Wuhan University,Wuhan 430072,China;Economic and Technology Research Institute of State Grid Fujian Electric Power Co.,Ltd.,Fuzhou 350013,China;Fujian Shuikou Power Generation Group Co.,Ltd.,Fuzhou 350004,China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2024年第22期171-181,共11页
Automation of Electric Power Systems
基金
国家电网公司科技项目(5108-202218280A-2-390-XG)。
关键词
电动船舶
充电负荷
场景生成
鲁棒性辅助分类
生成对抗网络
深度学习
electric ship
charging load
scenario generation
robust auxiliary classifier
generation adversarial network
deep learning