基于IA-BP算法的GPS高程拟合误差补偿方法

BP算法是校正GPS高程拟合误差的常用手段,但传统BP算法易陷入局部极小,使测量结果的精度稳定性差。提出一种基于免疫算法(IA)和BP神经网络结合的优化算法,利用IA进行全局搜索,然后,调用BP算法进行局部搜索。实验结果表明:该优化算法在训练多层前向神经网络时可有效地避免传统BP算法易陷入局部极小,并可保持其高预测精度,收敛速度快,具有寻优的全局性和精确性,进而提高了测量精度,且神经网络的GPS高程拟合误差与标准值间的相对误差均方差小于0.042 2,相对误差均值小于0.047 2,相对误差最大值小于0.050 3。

Regression for GPS height fitting based on IA-BP arithmetic

BP algorithm is widely used in calibrating measurement error of GPS height fitting.Conventional BP algorithm tends to get into infinitesimal locally,which worsens the stability of the measurement accuracy.An evolutionary neural network model based on IA-BP optimal algorithm is proposed.In the model,immene algorithm(IA)is first used for global search and then BP algorithm for local search.Experiment results indicate that the IA-BP optimal algorithm may effectively avoid getting into infinitesimal locally and have the merits of high prediction precision,rapid convergence,global superiority and accuracy for optimization.The results show that the root mean-square relative error is less than 0.042 2,the mean absolute relative error is less than 0.047 2,and the max absolute relative error is less than 0.050 3 between the predict value and the standard one.