用人工神经网络方法计算太阳光谱辐照度分布  

Simulation of solar spectral irradiance distribution with artificial neural network

作　　者：马志新 张雅婷[1] 刘文柱[1] 韩安军[1] 刘正新[1,2,3] MAZhixin;ZHANG Yating;LIU Wenzhu;HAN Anjun;LIU Zhengxin(Shanghai Institute of Microsystems and Information Technology,Chinese Academy of Sciences,Shanghai 200050,China;School of Physical Science and Technology,Shanghai Tech University,Shanghai 201210,China;Research Center for Materials and Optoelectronics,University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院上海微系统与信息技术研究所,上海200050 [2]上海科技大学物质科学与技术学院,上海201210 [3]中国科学院大学材料与光电研究中心,北京100049

出　　处：《电源技术》2023年第4期518-522,共5页Chinese Journal of Power Sources

基　　金：上海市科委南极行动专项(19dz1207602,20dz1207100)。

摘　　要：针对传统太阳光谱辐照度分布计算模型计算速度慢,需要大量气象参数等问题,提出了一种基于人工神经网络(ANN)算法模型计算特定气象参数下的太阳光谱辐照度分布的方法。首先将数据集通过Mini-Batch-Kmeans算法聚类并标签化,利用Python平台搭建ANN算法模型并对数据集进行训练,从而获得该气象条件下的太阳光谱辐照度分布。结果表明,相对于其他机器学习算法模型,提出的ANN算法具有更好的误差评估参数表现以及更稳定的预测性能。此预测模型可运用于光伏器件的输出性能参数的计算,为光伏组件实际应用中预测发电量输出和性能评估提供了便捷工具。Conventional simulation method for solar spectral irradiance distribution needs a large number of parameters and long time for calculation.To simplify the simulation process and adapt to practical application,this paper proposed an artificial neural network(ANN)algorithm model to calculate the solar spectral irradiance with meteorological parameters.The input data was first clustered and labelled by the Mini-Batch-Kmeans algorithm,an ANN algorithm model was built using a Python platform and the data was trained,then the solar spectrum irradiance distribution could be output through repeatedly training.The results show that the ANN algorithm proposed in this paper has good error evaluation and more stable prediction performance.This prediction model can be applied to the calculation of the output performance of real photovoltaic system,providing a convenient tool for the performance evaluation and prediction of photovoltaic modules in further applications.

关 键 词：人工神经网络 聚类算法 太阳光辐照强度 太阳光谱辐照度分布 气象参数 

分 类 号：TM914.4[电气工程—电力电子与电力传动]