极限学习机在湖库总磷、总氮浓度预测中的应用

Application of extreme learning machine to total phosphorus and total nitrogen forecast in lakes and reservoirs

基于传统BP人工神经网络模型训练速度慢、参数选择困难、易陷入局部极值等问题,提出极限学习机(ELM)的水质预测模型。以云南某水库为例,选取NH3-N、NO2--N、NO3--N、CODMn和水体透明度作为网络输入,TP、TN作为输出,构建基于ELM的湖库TP、TN预测模型,并将ELM预测结果与传统BP、GA-BP、RBF人工神经网络模型模拟结果进行比较。结果表明,ELM模型预测精度高于传统BP和RBF模型模拟结果,甚至略高于GA-BP模型的预测精度,并且ELM模型具有参数选择简便、训练速度快、不会陷入局部最优值等特点,有着较大的计算优势。

The traditional BP neural network model has the disadvantages of low training speed and difficulty in parameter selection, and it can easily fall into local extremum issues. In order to solve these problems, a water quality prediction model, the extreme learning machine （ELM）, is proposed. The model was applied to a reservoir in Yunnan Province. NH3-N, NO2-N, NO3-N, CODMn, and water transparency were selected as the network inputs, and TP and TN were selected as the outputs to build the TP and TN prediction model based on ELM. The predicted results of ELM were compared with those of the traditional BP, GA-BP, and RBF neural network models. The results show that the ELM model＇ s prediction accuracy was higher than those of the traditional BP and RBF neural network models, and even slightly higher than that of the GA-BP model. In addition, ELM model parameter selection is simple, has a capability for fast training, and is not likely to fall into local optimum values; it has a large computational advantage.