基于MEEMD-PE与CS-WNN模型的网络时延预测

针对网络控制系统诱导时延具有的随机性、非平稳性、非线性等特点，提出了一种基于改进的集总平均经验模态分解（modified ensemble empirical mode decomposition，MEEMD）-排列熵和布谷鸟搜索（cuckoo search，CS）优化的小波神经网络（wavelet neural network，WNN）时延预测算法。首先通过MEEMD对网络诱导时延序列进行处理，分别计算各模态的排列熵值，对复杂度相近的模态进行重组后得到新的子序列，从而达到降低建模复杂度和减少计算量的目的；然后利用CS算法优化的WNN预测新的子序列；最后叠加各子序列预测结果以获得时延序列的最终预测值。仿真表明，该算法具有较好的预测精度，能反映时延序列的总体趋势，可有效地降低异常值影响等优点。

Network delay prediction based on model of modified ensemble empirical mode decomposition-permutation entropy and cuckoo search-wavelet neural network

Aiming at the characteristics of randomness, instability and nonlinear induced delay of networked control system, a delay prediction algorithm of modified ensemble empirical mode decomposition-permutation entropy (MEEMD-PE) and wavelet neural network (WNN) optimized by cuckoo search (CS) is proposed. Firstly, MEEMD is used to decompose the original delay sequence, then the permutation entropy values of each sub sequence are calculated and recombination of new subsequences to reduce the non-stationary characteristics of the delay sequence and amount of calculation. And then predict the new subsequences by the WNN optimized by the CS algorithm. The final results of network delay are calculated by superimposing the prediction of the submodels. Simulation results show that the method has the advantages of better prediction accuracy, reflecting the overall trend of the delay sequence, and effectively reducing the impact of outliers.