基于递阶粒子群优化径向基函数人工神经网络的光性能监控

Investigation on Optical Performance Monitoring Based on Optimized Radial Basis Function Artificial Neural Networks

为解决差错反向传输神经网络在透明可重构光网络光性能监测中精度不足的问题,提出一种基于优化的径向基函数人工神经网络的光性能监测方案。在该方案中,以信号眼图参数为网络输入,以光信噪比、色散和偏振模色散为网络输出;采用二进制与十进制相结合编码的递阶粒子群方法,用适应度函数引导粒子向小规模和小误差方向运动,进行神经网络的结构与参数自适应优化;分别以不同光信噪比,不同色散和偏振模色散水平仿真信道中传输速率为40 Gb/s差分相移键控仿真信号,进行网络训练和测试,并将测试结果与相同情形下基于差错反向传输法神经网络的光性能监测结果进行比较。结果表明,所提方案在保有人工神经网络方案优点的基础上,有着更好的监测精度。

For the improvement of optical performance monitoring in transparent and reconfigurable optical networks using artificial neural networks trained with eye-diagram parameters, radial basis function artificial neural network models to simultaneously identify three separate impairments that can degrade optical channels, namely optical signal-to-noise ratio, chromatic dispersion, and polarization-mode dispersion, are developed. The neural networks are trained with the parameters derived from eye-diagram as inputs and the tested levels of concurrent impairment as outputs. They are optimized by hierarchical particle swarm optimization method. In the process of network optimization, the particle swarm inclines to small scales and small errors by choosing proper fitness functions. Finally, the prediction of levels of concurrent impairment drawn from the optimized models is realized by simulation experiments, and a better performance compared with those based on backward propagation artificial neural network models under the same testing circumstances is obtained.