基于量子粒子群神经网络的太阳黑子数预测磁

摘要

为了提高太阳黑子数的预测精度，论文提出了一种基于量子粒子群神经网络预测太阳黑子数的模型（QPSO-BP 网络）。首先基于前18个太阳周（1755～1953）的年均值，利用量子粒子群算法优化 BP 神经网络的权值和阀值，完成网络训练训；然后对第19太阳周（1954～2013）年均值进行预测，检验模型的预测能力。与普通 BP 神经网络预测的对比结果表明，该模型在逼近能力和预测精度两方面均有明显提高，从而表明基于量子粒子群优化的训练方法对于提高神经网络预测能力具有一定潜力。%In order to enhance the prediction accuracy of sunspot numbers ,a novel prediction model based on quantum-behaved particle swarm-based neural networks was proposed .First ,taking the sunspot annual averages of first 18 Solar cycle (1755 ~ 1953) as the training set ,the weights and threshold values of BP neural networks were adjusted by quantum-behaved particle swarm optimization (QPSO) ,and then the training process was completed .Secondly ,the sunspot annual averages of the 19th Solar cycle (1954 ~ 2013) were employed to verify the prediction ability of the proposed model .Experimental results showed the proposed model was obviously superior to the general BP neural networks in both approximation ability and pre -diction accuracy ,which revealed the training based on QPSO had a certain potential in enhancing the prediction ability of the general BP neural networks .