LED路灯光照强度自适应控制方法  

Adaptive control method for the illumination intensity of LED street lamp

作　　者：林海军[1] 赖小强[1] 兰浩[1] 汪鲁才[1] 

机构地区：[1]湖南师范大学工程与设计学院,长沙410081

出　　处：《电子测量与仪器学报》2016年第6期887-894,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(51205127);湖南省高校产业化培育项目(13CY003);湖南省教育厅优秀青年项目(14B106)资助

摘　　要：根据黄昏时分自然光照缓慢变弱的特征,提出了一种基于神经网络的LED路灯光照强度自适应控制方法,实现了LED路灯光照强度随自然光照强度变化自动调节。该方法以LED路灯光照测量输出电压与光照强度之间函数关系的单调递增性(即导数大于0)为先验知识,构造神经网络模型训练的约束条件,并利用Lagrange乘子法构建增广拉格朗日函数作为训练目标函数,给出了详细的训练算法,完成神经网络优化设计,提高了LED路灯光照强度自适应控制模型的泛化能力。仿真实验表明,这种基于约束条件的神经网络方法(CCNN),比传统的数据驱动训练方法(DINN,即仅利用数据样本训练神经网络)具有更好的泛化能力,模型误差更小;现场测试表明,黄昏时分采用这种CCNN方法的LED路灯节能最大超过20%。The existing control method for LED street lamp is simple,and but it doesn＇t meet the energy-saving requirement. In this paper,an adaptive control method for the illumination intensity of LED street lamp based on neural network is proposed by using the illumination character of the ambient light in the evening. In this method,the prior knowledge i. e. the function of the LED ＇s output voltage and its illumination intensity is monotone increasing is used to generate the constraint condition,and then the performance index of training a neural network with Lagrange multiple method is constructed. Furthermore,the detailed training algorithm of a neural network is given. The simulation experiment results show that the generalization ability of this proposed method based on the neural network with constraint conditions（ CCNN） is better than that of the conventional neural network with data induction（ DINN,i. e. training a neural network by only using the data samples,not the prior knowledge）,and it has less error. In additional,the testing results show that the maximum energy-saving of the LED street lamp with CCNN is more than 20%.

关 键 词：LED路灯 光照强度 自适应控制 神经网络 约束条件 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN383[自动化与计算机技术—控制科学与工程]