BP神经网络和遗传算法相结合在洞庭湖氨氮浓度预测中的应用

Application of Combination of Genetic Algorithm and Artificial Neural Network for Predicting NH3-N Concentration

BP算法具有寻优精确的特点,而遗传算法具有很强的宏观搜索能力和良好的全局优化性能。因此将遗传算法与BP神经网络相结合,训练时先用遗传算法进行寻优,将搜索范围缩小后,再利用BP神经网络来进行精确求解,可以达到全局寻优和快速高效的目的。设计了一种利用遗传算法优化BP神经网络权重的预测方法,并对洞庭湖氨氮浓度的预测进行了研究。结果表明,丰水期(9月份)数据分布比较均匀,遗传算法优化BP神经网络权重的预测方法的3种学习算法计算值与实际值接近,并优于BP神经网络的计算结果,说明该方法具有较好的预测效果。

BP learning algorithm is excellent with rapid speed and high precision.Genetic Algorithm is a global optimization algorithm and has strong global search ability.In order to optimize the performance,mix genetic algorithm,its main thought is to combine genetic algorithm with BP neural network,which realizes that BP algorithm is inserted into genetic algorithm in the form of an operator.A new coding scheme of genetic algorithm is proposed to optimize weights distribution of BP neural network,and the prediction of NH3—N concentration was estimated.The results show that the distribution of the data,which is measured in the period of highest flow,is uniform,so the prediction values approximate to the real concentration for three different learning algorithms,and are more accurate than BP network.