基于GA-PSO混合优化的BPNN车速预测方法

BP神经网络(BPNN)已经用于车速预测方面的研究.针对BPNN不同的初始权值和阈值会影响车速预测精度的问题,提出一种基于GA-PSO混合优化的BPNN车速预测方法.以北工大西门到百葛桥为研究路径,构建基于BPNN的车速预测模型;将遗传算法(GA)和粒子群算法(PSO)的寻优过程进行融合,通过逐次迭代取最优的方式确定BPNN的最优初始权值和阈值,以此设计基于GA-PSO混合优化的BPNN车速预测方法.最后,以所选路径为对象,利用基于GA-BPNN的预测法、基于PSO-BPNN的预测法,以及提出的方法对车速进行了实验预测.结果表明,相较于前两种车速预测改进方法,本文方法的平均车速预测误差分别降低了37.1%和24.1%,有效地提高了车速的预测精度.

A BPNN-based Speed Prediction Method with GA-PSO Optimization Algorithm

The BP neural network (BPNN) is used to research on the speed prediction. For the problem that the different initial weights and thresholds of the BPNN can influence the speed prediction accuracy, a speed prediction method based on the BPNN with GA- PSO optimization algorithm is proposed. A route from Beigongdaximen to Baigeqiao is selected as the research path; then the speed prediction model based on the BPNN is established. Based on the optimization process between the genetic algorithm (GA) and particle swarm optimization (PSO), the speed prediction method based on the BPNN with GA- PSO optimization algorithm is designed through the method that the optimal weights and thresholds of the BPNN are determined by using the iteratively optimal method. Finally, based on the selected route, the GA- BPNNbased, the PSO-BPNN-based and the proposed speed prediction methods are used to achieve speed prediction, respectively. The result indicates that, compared with the other two optimal methods on the BPNN, the average speed errors of the proposed method are reduced by 37.1% and 24.1%, respectively. The proposed method can effectively improve the accuracy of the speed prediction.