遗传算法优化组合模型在卫星钟差预报中的应用

Application of Genetic Algorithm Optimized Combination Model in Satellite Clock Error Prediction

为了提高导航卫星钟差预报精度,本文在单一BP神经网络模型基础上引入遗传算法(GA)与自回归(AR)模型,构建新的GA-BP-AR模型。该组合模型充分发挥GA在网络模型参数寻优中的优势及AR模型在残差修正中的优势。首先,通过GA-BP模型对原始钟差进行建模预报;其次,使用AR模型对预报残差进行建模并进行外推预报;最后,将GA-BP模型预报值与AR模型残差预报值相加得到最终钟差预报结果。使用2组卫星钟差数据对本文的组合模型效果进行检验,并将实验结果与单一的模型预报结果进行对比。结果表明,本文提出的组合预报模型的预报精度最高,平均精度在0.3 ns以内,验证了本文提出组合预报模型的优越性与适用性。

In order to improve the accuracy of clock error prediction of navigation satellite,this paper introduces genetic algorithm(GA)and auto regression(AR)model on the basis of single BP(back propagation)neural network model,and constructs a new GA-BP-AR model.The combined model gives full play to the advantages of GA in network model parameter optimization and AR model in residual correction.Firstly,the original clock error is modeled and predicted by GA-BP model.Secondly,the prediction residual error is modeled and extrapolated by AR model;Finally,the final clock error prediction result is obtained by adding the GA-BP model prediction value and the AR model residual prediction value.Two sets of satellite clock error data are used to test the effect of the combined model,and the experimental results are compared with the prediction results of a single model.The results show that the prediction accuracy of the combined forecasting model proposed in this paper is the highest,and the average accuracy is less than 0.3 ns,which verifies the superiority and applicability of the combined forecasting model proposed in this paper.