GAN-CNN-GRU在光伏最大功率点跟踪中的应用  

Application of GAN-CNN-GRU in photovoltaic maximum power point tracking

作　　者：王冉冉 高慧敏[2] Wang Ranran;Gao Huimin(School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou,Zhejiang 310018,China;School of Information Science and Engineering,Jiaxing University)

机构地区：[1]浙江理工大学计算机科学与技术学院,浙江杭州310018 [2]嘉兴学院信息科学与工程学院

出　　处：《计算机时代》2023年第6期119-123,共5页Computer Era

基　　金：嘉兴市公益性研究计划项目(2020AY10012)。

摘　　要：光伏阵列在复杂光照强度条件下,出现局部遮荫现象,导致输出特性曲线呈现多峰值状态,造成光伏系统能量转化率低下。本文提出一种基于生成对抗网络(GAN)、卷积神经网络(CNN)和门控循环单元(GRU)的混合预测模型,预先使用GAN对光伏发电原始环境数据进行数据扩充,再用CNN提取不同光伏太阳能板的环境数据特征,最后利用GRU进行最大功率点电压的预测。仿真结果表明,该混合预测模型可避免陷入局部极值的情况,可以有效提高复杂光照强度条件下的最大功率点跟踪效率。The local shading in photovoltaic(PV)arrays under complex light intensity conditions leads to a multi-peak state in the output characteristic curve,resulting in low energy conversion rate of PV systems.A hybrid prediction model based on GAN,CNN and GRU is proposed.Firstly,GAN is used for data expansion of the original environmental data of PV power generation,then CNN is used to extract the environmental data features of different PV solar panels,and finally GRU is used to predict the maximum power point voltage.The simulation results show that the hybrid prediction model can avoid the situation of falling into local extremes and can effectively improve the maximum power point tracking efficiency under complex light intensity conditions.

关 键 词：光伏电池 最大功率点跟踪 GRU神经网络 CNN神经网络 生成对抗网络 仿真 

分 类 号：TM615[电气工程—电力系统及自动化]