Optimum coagulant forecasting by modeling jar test experiments using ANNs.
Optimum coagulant forecasting by modeling jar test experiments using ANNs.
复制标题
DOI：
10.5194/dwes-11-1-2018
复制
发表时间：
2018-01-01
影响因子：
--
通讯作者：
Moharramzadeh, S.
中科院分区：
文献类型：
--
作者：
HaghIrI, S.;Daghighi, A.;Moharramzadeh, S.
目前，水处理厂的适当利用和优化其使用特别重要。水处理中的混凝和絮凝是使用混凝剂导致颗粒不稳定和形成更大更重颗粒的常见方式，从而改善沉淀和过滤过程。确定这种促凝剂的最佳剂量具有特别重要的意义。除了增加成本之外，高剂量还可能导致沉淀物保留在滤液中，这是根据标准的危险状况，而不足剂量的凝结剂可能导致降低凝结过程所需的质量和可接受的性能。尽管瓶试验用于测试混凝剂，但是这样的实验在评估由输入水的突然变化产生的结果方面面临许多限制，因为它们的显著成本、长时间要求以及许多因素（浊度、温度、pH、碱度等）之间的复杂关系。这可能会影响混凝剂的效率和测试结果。建模可以用来克服这些局限性，在这项研究中，人工神经网络（ANN）的多层感知器（MLP）与一个隐藏层已被用于模拟罐测试，以确定在水处理过程中使用的混凝剂的剂量水平。本研究中包含的数据来自伊朗阿尔达比尔省的饮用水处理厂。为了评估模型的性能，均方误差（MSE）和相关系数（R2）参数已被使用。所获得的值是在一个可接受的范围内，证明了高精度的模型相对于水质特性和混凝剂的最佳剂量的估计，所以使用这些模型将允许运营商不仅减少成本和时间进行实验罐测试，但也预测一个适当的剂量混凝剂的量和项目的质量输出水在真实的条件下。
Currently, the proper utilization of water treatment plants and optimizing their use is of particular importance. Coagulation and flocculation in water treatment are the common ways through which the use of coagulants leads to instability of particles and the formation of larger and heavier particles, resulting in improvement of sedimentation and filtration processes. Determination of the optimum dose of such a coagulant is of particular significance. A high dose, in addition to adding costs, can cause the sediment to remain in the filtrate, a dangerous condition according to the standards, while a sub-adequate dose of coagulants can result in the reducing the required quality and acceptable performance of the coagulation process. Although jar tests are used for testing coagulants, such experiments face many constraints with respect to evaluating the results produced by sudden changes in input water because of their significant costs, long time requirements, and complex relationships among the many factors (turbidity, temperature, pH, alkalinity, etc.) that can influence the efficiency of coagulant and test results. Modeling can be used to overcome these limitations; in this research study, an artificial neural network (ANN) multi-layer perceptron (MLP) with one hidden layer has been used for modeling the jar test to determine the dosage level of used coagulant in water treatment processes. The data contained in this research have been obtained from the drinking water treatment plant located in Ardabil province in Iran. To evaluate the performance of the model, the mean squared error (MSE) and correlation coefficient (R2) parameters have been used. The obtained values are within an acceptable range that demonstrates the high accuracy of the models with respect to the estimation of water-quality characteristics and the optimal dosages of coagulants; so using these models will allow operators to not only reduce costs and time taken to perform experimental jar tests but also to predict a proper dosage for coagulant amounts and to project the quality of the output water under real conditions.