基于天牛群-反馈神经网络的绞线串扰预估

在三绞线间的串扰问题中,绞线不同扭转角度会带来单位长度(per unit length,PUL)RLCG寄生参数的变化,已不能通过常规方法直接求解传输线方程得到串扰. 需借助于频域链参数的理论,将三绞线进行若干分段,通过对每一段级联得到串扰. 提出天牛群(beetle swarm optimization,BSO)算法优化反馈神经网络(back propagation neural network,BPNN)的权值,使其误差更小. 预测绞线不同分段截面处的寄生参数,并将BSO算法与天牛须(beetle antennae search,BAS)算法的寻优能力进行比较. 最后,基于BSO-BP、BAS-BP和BP共3种方法所预测的寄生参数进行串扰求解,并与CST cable studio仿真值进行比较验证. 结果表明,BSO-BP算法与仿真值相比具有最好的吻合度,而初始的BP算法则效果最差.

Prediction of Twisted Wire Crosstalk Based on Beetle Swarm Optimization-Back Propagation Neural Network

With the development of current power equipment toward miniaturization,high frequency,and high power,the crosstalk caused by electromagnetic coupling between adjacent cables has become a problem that cannot be ignored. The research object of this paper is crosstalk between triple twisted wires. Different twist angles of twisted wires will bring about the change of per unit length(per unit length,PUL)RLCG parasitic parameters. It is no longer possible to directly solve the transmission line equation and obtain the crosstalk with conventional methods. With the help of the theory of frequency domain chain parameters,triple twisted wires are divided into several sections,and crosstalk is finally obtained by cascading each section. This paper proposes the Beetle Swarm Optimization(BSO)algorithm to optimize the weights of Back Propagation Neural Network(BPNN)to make the error smaller. Parasitic parameters at different sections of the strands are predicted,and optimization capabilities of BSO algorithm with Beetle Antennae Search(BAS)algorithm are compared. Finally,crosstalk is obtained on the basis of the parasitic parameters predicted by three methods of BSO-BP,BAS-BP and BP. Furthermore,crosstalk are compared with the simulation value of CST cable studio. The results show that BSO-BP algorithm has best agreement with simulated value,while the initial BP algorithm has the worst effect.