生物脱氮除磷活性污泥系统复合模拟方法

为避免繁琐的参数校核工作,提出了活性污泥2 d号模型(ASM2d)和人工神经网络(ANNs)相结合的复合模拟方法。考察了复合方法在某污水处理厂生物脱氮除磷工艺中的应用情况。研究表明,ANNs能够准确地模拟出水实测值与未经校核的ASM2d机理模型的估计值之间的差值。利用Levenberg-Marquardt算法,对出水氨氮、总氮和总磷分别建立网络结构为5-12-1、5-8-1和5-8-1的ANNs子模型,将这些子模型输出同ASM2d机理模型输出相加便得到复合模型输出。复合模型估计值对前10.4 d(ANNs子模型训练数据时段)出水氨氮、总氮和总磷浓度的拟合平均绝对百分比误差分别为0.267、0.055和0.048;其对后2.6 d(ANNs子模型测试数据时段)出水氨氮、总氮和总磷浓度的预测平均绝对百分比误差分别为0.332、0.083和0.069。均方根误差、平均绝对误差等评价指标也表明复合模型能够给出合理的模拟结果。

Modeling biological phosphorus and nitrogen removal in activated sludge systems: A hybrid approach

This paper studied a hybrid approach combining the activated sludge model No. 2d (ASM2d) and an artificial neural network (ANN). The hybrid approach was evaluated using 13 day measurements of a full scale activated sludge process. Results demonstrated that ANNs were able to simulate the difference between the measurements and estimates of an uncalibrated ASM2d. Based on Levenberg-Marquardt algorithm, three back-propagation ANN sub-models were established by trial and error with the structures of 5-12-1, 5-8-1 and 5-8-1 for effluent ammonia nitrogen (NH⁺₄-N), total nitrogen (TN) and total phosphorus (TP), respectively. Then, outputs of the ANN sub-models were added to estimates of the ASM2d, which led to estimates of a hybrid model. The hybrid model fitted the measurements of NH⁺₄-N, TN and TP with mean absolute percentage errors (MAPEs) of 0.267, 0..055 and 0.048, respectively, for the data of the first 10.4 days, which acted as training data for ANN sub-models. For the remained data, which acted as testing data for ANN sub-models, the hybrid model estimated the effluent concentrations of NH⁺₄-N, TN and TP with mean absolute percentage errors (MAPEs) of 0.332, 0.083 and 0.069. Moreover, the hybrid model offered satisfactory performance by evaluation of root mean square errors and mean absolute errors.