Elman神经网络在区域速度场建模中的应用

鉴于欧拉矢量参数在局部区域适应性较差，提出一种基于Elman神经网络的速度场逼近方法。首先利用已有欧拉矢量参数估计站点速度，将剩余残差作为Elman神经网络拟合量进行逼近；然后将Elman神经网络估计结果与欧拉矢量计算速度相叠加，获得区域速度场模型。利用山东区域速度场数据进行验证，结果表明，该方法在一定程度上能够削弱系统误差影响，提高计算精度。

Application of Elman Neural Network in Velocity of Local Area

The parameter of Euler vector has poor validity in a local area. An algorithm based on Elman neural network is introduced to fit the velocity field. First, the velocity of the position is computed with the parameter of Euler vector; second, the residual as the expectation of Elman neural network is again trained; finally, we determine the velocity of the position, which is equal to the sum of the results obtained by Elman neural network and the velocity of Euler vector. The data set of Shandong is employed to test the algorithm. It is shown that the new algorithm can weaken the influence of systemic error and improve the accuracy of velocity field.