说明：  对具有随机噪声的二阶系统的模型辨识，进行标幺化以后系统的参考模型差分方程为： y(k)=a1*y(k-1)+a2*y(k-2)+b*u(k-1)+s(k) 式中，a1=0.3366,a2=0.6634,b=0.68,s(k)为随机噪声。由于神经网络的输出最大为1，所以，被辨识的系统应先标幺化，这里标幺化系数为5。采用正向建模(并联辨识)结构，神经网络选用3-9-9-1型，即输入层i，隐层j包括2级，输出层k的节点个数分别为3、9、9、1个；由于神经网络的最大输出为1，因此在辨识前应对原系统参考模型标么化处理，辨识结束后再乘以标么化系数才是被辨识系统的辨识结果。
(of random noise with the second-order system model, per-unit system after the reference model differential equation : y (k) = y* a1 (k-1) a2* y (k-2) b* u (k-1) s (k)- style, = 0.3366 a1, a2 = 0.6634, b = 0.68, s (k) as random noise. Because the neural network for a maximum output, therefore, the identification system should be per-unit, per-unit here coefficient of 5. Forward modeling (Parallel identification) structure, neural network-based selection 3-9-9-1, i input layer, hidden layer, including two j, k output layer to the number of nodes 3,9,9,1; The neural network the biggest losers up to one, in the original deal before Identification System Reference Model S Mody treatment, then multiplied by the end of Identification Standard Mody coefficient was recognition system is the ide)