基于改进支持向量回归机的污水处理厂出水总氮预测模型

Prediction of effluent total nitrogen concentration in a wastewater treatment plant using a particle swarm optimization-support vector regression model

在小样本数据的情况下,采用粒子群优化算法(PSO)对传统支持向量回归机(SVR)进行改进,将其应用于北京某大型污水处理厂出水总氮浓度预测上。预测结果精度对比分析表明,PSO-SVR模型预测结果平均相对误差为1.836%,决定系数为67.76%,均方根误差为0.693 9,各评价指标均优于多元线性回归模型、BP神经网络模型。因此在小样本情况下,利用PSO-SVR模型对污水处理厂出水总氮浓度进行预测是可行有效的,为应用数据驱动模型对污水处理过程进行建模模拟提供了一种新方法尝试。

A particle swarm optimization （PSO）-support vector regression （SVR） was built based on small sample and applied it to predict effluent total nitrogen concentration in a wastewater treatment plant. The analysis of prediction accuracies indicated that the mean relative error （MRE） is 1. 836% , the coefficient of determina- tion （R2） is 67.76% as well as the root mean square error （RMSE） is 0. 693 9. In addition, the accuracy of the PSO-SVR model was analyzed by comparison with the multivariable linear regression （MLR） model and the BP neural network （BP-ANN）. The results indicated that the PSO-SVR model is better than MLR and BP-ANN in prediction of effluent total nitrogen concentration in a wastewater treatment plant. Therefore, it is feasible and effective to predict effluent total nitrogen concentration in a wastewater treatment plant by using PSO-SVR model, which provides the method to modeling the process of wastewater treatment.