基于GA-GRU神经网络的光伏MPPT算法  

MPPT ALGORITHM FOR PHOTOVOLTAICS BASED ON GA-GRU NEURAL NETWORK

作　　者：王冉冉 高慧敏[2] 张昕宇 Wang Ranran;Gao Huimin;Zhang Xinyu(School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310018,China;College of Information Science and Engineering,Jiaxing University,Jiaxing 314001,China;Zhejiang Jinko Solar Co.,Ltd.,Haining 314400,China)

机构地区：[1]浙江理工大学计算机科学与技术学院,杭州310018 [2]嘉兴学院信息科学与工程学院,嘉兴314001 [3]浙江晶科能源有限公司,海宁314400

出　　处：《太阳能学报》2023年第9期212-219,共8页Acta Energiae Solaris Sinica

基　　金：嘉兴市公益性研究计划(2020AY10012)。

摘　　要：针对外界环境因素快速变化时,光伏发电系统难以保持在最大功率点输出的问题,提出遗传算法与GRU神经网络相结合的最大功率跟踪算法(GA-GRU-MPPT)。该算法在构建的最大功率点预测模型基础上,采用遗传算法对GRU神经网络的参数进行优化。考虑到数据的关联性,将前一时刻的太阳电池温度、太阳辐照度、最大功率点电压及当前时刻的太阳电池温度和太阳辐照度作为预测模型的输入变量,输出为当前时刻的最大功率点电压。针对3种不同气候情形的仿真结果表明,该算法跟踪精度可达99%,能显著提高光伏系统的能量转换效率。This paper proposes a maximum power point tracking algorithm that combines genetic algorithm and GRU neural network(GA-GRU-MPPT)to address the problem of photovoltaic power generation systems struggling to maintain maximum power point output in the face of rapidly changing external environmental factors.Based on the constructed maximum power point prediction model,this al⁃gorithm optimizes the parameters of the GRU neural network using genetic algorithm.Considering the correlation of the data,the previ⁃ous moment's solar cell temperature,solar irradiance,maximum power point voltage,as well as the current moment’s solar cell temper⁃ature and solar irradiance,are taken as input variables for the prediction model,with the output being the current moment’s maximum power point voltage.Simulation results for three different climate scenarios show that the tracking accuracy of this algorithm can reach 99%,significantly improving the energy conversion efficiency of solar photovoltaic systems.

关 键 词：太阳电池 最大功率点跟踪 遗传算法 GRU神经网络 仿真 

分 类 号：TM615[电气工程—电力系统及自动化]