摘要
提出一种把人工神经网络和光学仿真软件相结合的太阳模拟器光均匀性优化设计方法。首先由光学软件仿真一组不同结构参数的模型,然后让神经网络通过学习这些仿真数据来寻找模型的结构参数和检测平面辐照均匀性的复杂非线性关系,并预测辐照均匀性最优时的模型结构参数,再用仿真软件进行验证。基于此方法,对一个短距离大面积的太阳模拟器模型成功进行了优化仿真,使其辐照非均匀度满足A级标准。结果表明:此方法能有效解决太阳模拟器设计周期长、调试困难、试验成本高的问题,对解决同类工程问题亦具有一定的研究价值。
A new optimal design method of the irradiance uniformity of solar simulator based on artificial neural network was proposed. Firstly, the learning materiel for neural network can be got by simulating some models with TracePro. And then the neural network can be used to forecasting the optimal model. A diffuse reflecting solar simulator box has been designed successfully, which has a low height and large test area and its irradiance non-uniformity is less than 2%. The result indicates that it can shorten the design period and cut the experimental cost by this way. The method can be also applied to solve the similar engineering problems.
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2009年第9期1177-1181,共5页
Acta Energiae Solaris Sinica
基金
上海市科委重点科技攻关项目(05DZ0312)