粒子群优化的BP神经网络在电梯群控系统中的应用

目的提出基于粒子群优化的BP神经网络获取评价电梯群控系统派梯性能指标的新方法．方法综合考虑电梯运行特性，确定电梯调度控制策略，建立了电梯运行性能的评价指标函数，利用神经网络自学习功能获取评价指标的初始权值和阀值，针对平均候梯时间对比研究了普通BP神经网络算法和粒子群优化BP神经网络算法．结果将优化的权值和阀值代入BP神经网络获得平均候梯时间，粒子群优化的BP神经网络与BP神经网络相比，减少了迭代次数，缩短了运行时间．结论仿真实验表明，该方法可以避免BP神经网络训练中产生局部极小值，加快BP神经网络训练速率，提高电梯群控系统控制的速度．

Application of BP Neural Network Optimized by Particle Swarm Optimization to Elevator Group Control System

In this paper, a new method to obtain evaluation index of the elevator group control system （EGCS） was proposed applying BP neural network based on Particle Swarm Optimization algorithm. The operation characteristics of elevators were considered comprehensively, the control strategy of elevators was determined and evaluation index function of elevators was established. The weights and biases of evaluation index function were obtained by using learning function of BP neural network. According to average waiting times, the two algorithms of BP neural network and BP neural network based on Particle Swarm Optimization were comparatively researched. The weights and biases were substituted in the BP neural network of elevator group control system to obtain average waiting times. The simulation results show that BP neural network based on Particle Swarm Optimization algorithm can reduce training numbers and times, and can avoid local minimum of BP neural network,improve the rate of BP neural network and the speed of optimal control of elevator group control system.