基于免疫进化算法的BP网络模型在径流预测中的应用

在BP网络模型基础上,引入免疫进化算法,建立了基于免疫进化算法的BP网络模型。该模型在一定范围内随机生成初始权值群体,用BP网络进行训练,选择群体中具有最大适应度值的权值个体作为最优个体,应用免疫进化算法生成下一代权值群体,再用BP网络对权值群体进行训练,此迭代过程反复进行,直至达到问题求解精度要求为止。将该模型应用于新疆伊犁河雅马渡站的径流预测,并将预测结果与基本BP网络模型和基于遗传算法的模糊优选BP网络模型的预测结果进行比较,结果表明,该模型不仅精度较高,而且稳定性较好。

Runoff prediction application of BP neural network model based on immune evolutionary algorithm

A kind of BP neural network model was established based on the immune evolutionary algorithm （IEABP）. The initial weights group was randomly generated within a certain range. The BP neural network model then trained the initial weights group. From the weights group, the best fitness weight individual was selected as the optimal weight individual and then expanded to the same dimensions as the next generation weights group with the immune evolutionary algorithm. The renewed weights group was then trained with the BP neural network model again. This procedure was repeated until the pre-set precision of the problem was reached. The IEABP mixed network model was used to forecast annual runoff at Yamadu Station in Yili, Xinjiang. Compared with the results from a general BP neural network model and the fuzzy optimization approach based on a genetic algorithm, this model is more precise and more stable.