摘要
针对传统照度计算中利用系数计算过程繁琐、误差大的问题,提出并实现了基于Adam优化算法的由固定网络和可变网络并联构成的神经网络模型,进行灯具的利用系数拟合计算,分别拟合了计算地板反射比为0.2时的利用系数和地板反射比不为0.2时利用系数修正系数。使用训练好的模型代替传统的利用系数查表过程,降低了照度计算的计算误差,提高了工程实用性。实验结果表明,最大误差率约为2%。
Aiming to the phenomenon that the calculation of utilization factor is complicated and not accurate in traditional illumination calculation, a neural network model optimized by Adam algorithm and consisting of a fixed network and a variable network in parallel is designed and realized to fit the coefficient calculation. The utilization factor when the floor reflectance is 0.2 and the utilization factor correction factor when floor reflectance is not 0.2 are calculated separately. Replacing the traditional look-up process with trained model can reduce calculation error, and improve the engineering practicability. The experimental results show that the maximum error rate is about 2%, far less than the requirements in the standard.
作者
汤烨
陆卫忠
陈成
黄宏梅
TANG Ye;LU Weizhong;CHEN Cheng;HUANG Hongmei(School of Electronic and Information Engineering, Suzhou University of Science and Technology, Suzhou 215009, China;Jiangsu Key Laboratory of Intelligent Building Energy Efficiency, Suzhou 215009, China;Virtual Reality Key Laboratory of Intelligent Interaction and Application Technology of Suzhou, Suzhou 215009, China)
出处
《照明工程学报》
2019年第2期50-54,共5页
China Illuminating Engineering Journal
基金
国家自然科学基金项目(批准号:61672371)
江苏省教育厅自然科学研究项目(批准号:08KJD510007)
苏州市科技发展计划重点实验室项目(批准号:SZS201609)
关键词
Adam算法
神经网络
照度计算
利用系数
Adam algorithm
neural network
illumination calculation
utilization factor
