论文部分内容阅读
网络控制系统中的时延是影响系统性能的重要参数,针对基于Internet的网络控制系统中时延预测问题,提出一种最大Lyapunov指数与Elman神经网络结合的预测方法.首先对时延序列进行相空间重构,得到嵌入维数与延迟变量,然后通过最大Lyapunov指数方法与Elman神经网络对时延分别进行一步预测,将两种预测方法的预测结果通过不同的权值系数进行叠加得到最终的时延预测值.最后针对权值系数的寻优问题,提出一种改进的自由搜索算法,其收敛精度与速度都优于标准的自由搜索算法.仿真实验表明,相对于其它预测方法,本文的基于Lyapunov-Elman的时延预测方法具有较高的预测精度与较小的预测误差.
Time delay in networked control system is an important parameter that affects system performance. Aiming at the problem of delay prediction in networked control system based on Internet, a prediction method combining maximum Lyapunov exponent with Elman neural network is proposed.Firstly, Space reconstruction, the embedding dimension and delay variable are obtained, and then the maximum Lyapunov exponent method and the Elman neural network are used to predict the delay separately. The prediction results of the two prediction methods are superposed by different weight coefficients to obtain the final time Delay prediction.Finally, aiming at the optimization of weight coefficient, an improved free-search algorithm is proposed, whose convergence accuracy and speed are better than the standard free-search algorithm.The simulation results show that compared with other prediction methods, Lyapunov-Elman’s delay prediction method has higher prediction accuracy and smaller prediction error.