MEA-Elman与PSO-Elman预测电离层f0F2对比研究

Comparison of MEA-Elman and PSO-Elman for Prediction of Ionosphere fF

针对Elman神经网络对电离层临界频率f0F2预测的不足的问题,引入思维进化算法和粒子群算法优化Elman神经网络。首先,根据f0F2预测要求确定Elman神经网络各层节点个数。然后,分别利用思维进化算法和粒子群算法优化Elman神经网络的初始权值和阈值。最后,分析了两种优化算法对f0F2的预测性能。通过将预测值和实际值比较,验证了优化算法的精确度。结果表明,PSO-Elman和MEA-Elman算法对f0F2的预测精度明显大于Elman神经网络预测精度,当上午的f0F2值较大时,MEA-Elman算法表现出的连续数据预测性能优于PSO-Elman算法。

In view of the shortcomings of Elman neural network for the prediction of ionospheric critical frequency f0F2, this paper introduces thought evolution algorithm and particle swarm optimization algorithm to optimize Elman neural network. First, we determined the number of nodes in each layer of the Elman neural network according to the f0F2 prediction requirement. Then, we used the thought evolution algorithm and particle swarm optimization algorithm to optimize the initial weight and threshold of Elman neural network. Finally, the prediction performance of two optimization algorithms for f0F2 was analyzed. This paper verified the accuracy of the optimization algorithm by comparing the predicted value with the actual value. The results show that the prediction accuracy of the PSO-Elman and MEA-Elman algorithms for f0F2 is significantly greater than that of the Elman neural network. When the value of f0F2 in the morning is large, the MEA-Elman algorithm exhibits better continuous data prediction performance than the PSO-Elman algorithm.