基于光效和BP神经网络的LED结温预测研究  

Research on LED Junction Temperature Prediction Based on Luminous Efficiency and BP Neural Network

作　　者：王宸 张军朝[1,3,4,5,6] 许并社 张婕[1,3,4,5,6] 付强 WANG Chen;ZHANG Junchao;XU Bingshe;ZHANG Jie;FU Qiang(College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan Shanxi 030024,China;MOE Key Laboratory of Interface Science and Engineering in Advanced Materials,Taiyuan University of Technology,Taiyuan Shanxi 030024,China;Shanxi Electric Drive and Internet of Thing Engineering Research Center,Taiyuan Shanxi 030024,China;Shanxi Province“1331 Project”—Smart City Based on Big Dada Lighting Data Sharing and Public Service Engineering Technology Research Center,Taiyuan Shanxi 030024,China;Shanxi Intelligent Digital Lighting Union Laboratory,Taiyuan Shanxi 030024,China;Shanxi Intelligent Digital Cultural Tourism Industry Research Institute,Taiyuan Shanxi 030024,China)

机构地区：[1]太原理工大学电气与动力工程学院,山西太原030024 [2]太原理工大学新材料界面科学与工程教育部重点实验室,山西太原030024 [3]山西省电气传动及物联网工程研究中心,山西太原030024 [4]山西省“1331工程”基于大数据的智慧城市照明数据共享与公共服务平台工程技术研究中心,山西太原030024 [5]山西省智能照明联合实验室,山西太原030024 [6]山西省智能数字文化旅游产业研究院,山西太原030024

出　　处：《电子器件》2024年第2期496-501,共6页Chinese Journal of Electron Devices

基　　金：山西省“1331工程”基于大数据的智慧城市照明数据共享与公共服务平台工程技术研究中心专项建设项目(SC19100026);山西省电气传动及物联网工程研究中心建设项目(RD1900000333);山西省研究生教改项目(2017JG25)。

摘　　要：提出了一种发光效率结合BP神经网络的大功率LED结温预测方法。实验中发现LED的发光效率在结温升高至80℃左右将急剧下降,光效与结温的函数关系发生改变,进而影响了测量精度。针对这一问题,基于发光效率与结温的函数关系,构建实验平台,获取发光效率和相应结温的数据,然后通过BP神经网络建立LED结温预测模型。模型所得数据与正向电压法进行对比实验,最大误差为2.1℃,验证了所提方法的可行性,同时所提方法无需考虑LED的内部结构,能够简便,准确地预测大功率LED结温。A prediction method of junction temperature of high⁃power LED based on luminous efficiency and BP neural network is pro⁃posed.It is found that the luminous efficiency of LED will drop sharply when the junction temperature rises to about 80℃,and the functional relationship between luminous efficiency and junction temperature will change,which will affect the measurement accuracy.To solve this problem,based on the functional relationship between luminous efficiency and junction temperature,an experimental plat⁃form is constructed to obtain the data of luminous efficiency and corresponding junction temperature,and then the LED junction temper⁃ature prediction model is established through BP neural network.The maximum error between the data obtained from the model and the forward voltage method is 2.1℃,which verifies the feasibility of the proposed method.At the same time,the internal structure of LED does not need to be considered,and the junction temperature of high⁃power LED can be easily and accurately predicted.

关 键 词：大功率LED 发光效率 BP神经网络 结温 

分 类 号：TM930.1[电气工程—电力电子与电力传动]