摘要
针对BP神经网络在处理非线性与复杂环境时易陷入局部最优解,且收敛速度较慢的问题,本文提出一种改进的人工蜂群(ABC)算法用于优化反向传播神经网络模型,并将其应用于钟差短期预报中。首先,从增强步长的随机性、提升搜索效率及维持种群多样性出发,引入莱维飞行策略、教与学优化算法及适应度-距离平衡机制,改进ABC算法,有效提高算法的全局搜索能力,避免陷入局部最优解。其次,将改进的ABC算法与BP神经网络相结合,应用于卫星钟差短期预报,并给出相应的计算步骤。然后,选用GFZ提供的高精度卫星钟差产品,从算法效率、稳定性及精度等方面进行单天和多天预报对比分析,验证模型的适用性。最后,验证优化模型的预报结果在PPP中的应用效果。结果表明,改进后的人工蜂群算法能够快速逼近最优解,精度提升明显,与二次多项式(quadratic polynomial,QP)模型、BP和ABC-BP模型相比,平均精度分别提升了56.55%、25.11%和7.07%,且MEO-PHM钟的提升效果优于MEO-Rb钟;改进人工蜂群算法与BP组合模型残差分布更加集中,中位数更接近零,极值更小,在6、12 h的预报中具有较高的准确性和稳定性;使用预报钟差序列进行精密单点定位测试,改进组合模型的结果在E、N、U方向的精度,分别较ABC-BP模型和BP模型提升了42.07%、31.07%,41.79%和45.42%、50.16%、46.18%。
In order to solve the problem that BP neural network is easy to fall into the local optimal solution and the convergence speed is slow when dealing with nonlinear and complex environments,an improved artificial bee colony algorithm is proposed to optimize BP neural network and apply it to short-term prediction of clock deviation.Firstly,from the perspective of enhancing the randomness of the step size,improving the search efficiency and maintaining the diversity of the population,the Lévy flight strategy,the teaching and learning optimization algorithm and the fitness-distance balance mechanism were introduced to improve the artificial bee colony algorithm,which effectively improved the global search ability of the algorithm and avoided falling into the local optimal solution.Secondly,the improved artificial bee colony algorithm is combined with BP neural network to be applied to the short-term prediction of satellite clock deviation,and the corresponding calculation steps are given.Then,the high-precision satellite clock products provided by GFZ are selected to compare and analyze the single-day and multi-day forecasts from the aspects of algorithm efficiency,stability and accuracy,so as to verify the applicability of the model.Finally,the prediction results of the optimization model are verified to be applied in PPP.Compared with the QP,BP and ABC-BP models,the average accuracy is increased by 56.55%,25.11%and 7.07%,respectively,and the improvement effect of MEO-PHM clock is better than that of MEO-Rb clock.The improved artificial bee colony algorithm and the BP combined model have a more concentrated residual distribution,a median closer to zero,and a smaller extreme value,which has high accuracy and stability in the 6 and 12 h prediction.The accuracy of the combined model in the E,N and U directions was improved by using the forecast clock error sequence to improve the accuracy of the results in the E,N and U directions,which were improved by 42.07%,31.07%,41.79%and 45.42%,50.16%,46.18%compared with the ABC-BP model and the BP model,respectively.
作者
潘雄
张龙杰
艾青松
金丽宏
蔡茂
PAN Xiong;ZHANG Longjie;AI Qingsong;JIN Lihong;CAI Mao(School of Computer Science and Artificial Intelligence,Wuhan Textile University,Wuhan 430200,China;School of Mathematics and statistics,Wuhan Textile University,Wuhan 430200,China;Changjiang Space Information Technology Engineering Co.,Ltd.,(Wuhan),Wuhan 430010,China)
出处
《测绘学报》
北大核心
2025年第8期1404-1415,共12页
Acta Geodaetica et Cartographica Sinica
基金
国家自然科学基金(42174010,41874009)
湖北省自然科学基金(2023AFB435)。
关键词
适应度-距离平衡
教与学优化
短期预报
精密单点定位
fitness-distance balance
teaching-learning-based optimization
short-term prediction
precision point positioning
