基于BP神经网络的水蒸气红外透过率仿真

针对在计算大气透过率方法中利用经验公式误 差大和专业软件复杂低效等问题,提出了一种基于 贝叶斯正则化BP(back propagation)神经网络的水蒸气红外透过率的计算方法。利用BP 神 经网络良好的非线性拟合特点,建 立大气参数与水蒸气透过率之间的关系模型。以实测温度、压强和湿度作为输入向量,中红 外平均水蒸气 透过率作为输出,构建3-7-1式的BP神经网络。仿真结果表明: 在相同的大气参数下,与逐线积分法相比, 本文方法在运算过程大幅简化的同时相对误差很小；与经验公式法相比,本文方法对透过率 的计算精度大幅提升。

Simulations of infrared water vapor transmittance based on BP meural network

Because there are large errors using empirical formula and great complexity usin g software in methods of calculating water vapor transmittance,a method based on back propagation (BP) neural network with Beyesian regularization is proposed.The model of relationship between atm ospheric data and infrared water vapor transmittance is built by using the characteristic of nonlinear fitting of BP neural network.The input factors include measured temperature,pressure and relative humidity.The output variable is average mid-infrared water vapor transmittance. The 3-7-1type BP neural network model is built.Simulations in dicate that the process of calculation is simplified a lot and the relative errors are very small compared with those in t he method of line-by-line. The precision of transmittance is improved a lot by this methods compared with t hat by empirical formula. Therefore,the methods provide helpful lesson to calculate atmospheric transmitt ance fast and accurately.