Wastewater treatment is a critical step for water environment quality promotion and pollutant reduction. The wastewater treatment technology, influent shock, and complexity of natural conditions lead to variation and uncertainty in the wastewater treatment process. These uncertainties cause fluctuations in operation costs and effluent water quality.
Image Credit: Bilanol/Shutterstock.com
Artificial intelligence (AI) has become an effective tool for minimizing wastewater treatment complexities. AI technology, specifically machine learning, is increasingly being applied to enhance the wastewater treatment plant capacity. This article deliberates on the role of AI in wastewater treatment.
Limitations of Conventional Approaches
The wastewater composition is extremely complex, with pollutant concentrations, influent properties, and treated effluent greatly varying across wastewater treatment plants. Wastewater treatment is a complex process that is affected by many microbiological, physical, and chemical factors. The influent variability and stochastic perturbations necessitate proper operational controls on the system.
Conventional wastewater treatment approaches depend on complex mathematical formulas and calculations of operation, process, and design parameters, along with technical, governmental, regulatory, social, and cost constraints. Multivariate and statistical data analysis methods have been utilized to evaluate the pollutants' removal efficiencies and estimate uncertainties.
Importance of AI
AI technologies can effectively handle dynamic, interactive, and complex wastewater treatment problems. They provide more accuracy in the prediction of regression coefficients and significantly less average error without using detailed information regarding the relationship between the output and input variables and complex mathematical formulas.
Actions required to prevent the effluent quality deterioration can be taken in time by employing AI technologies to predict the treatment process’ effluent quality. Automated sampling assists the operator in understanding the patterns and trends and prevents disasters through timely action.
Additionally, constant monitoring of systems using sensors improves operational efficiency, reduces costs, and optimizes energy consumption. Hybrid AI and statistical tools have been initially utilized to improve wastewater treatment plant operations. Later, efforts have been made to apply AI in predicting influent, effluent, aeration time, energy consumption, and filamentous sludge bulking in wastewater treatment.
Different AI model types like hybrid models and neural networks have been developed to understand the nonlinear pollutant removal pattern in the treatment processes. Specific parameters like biological oxygen demand (BOD), chemical oxygen demand (COD), emerging pollutants, and heavy metals have been the focus of several AI model-based studies to evaluate wastewater treatment performance.
AI Techniques in Wastewater Treatment
Artificial neural network (ANN), neural-fuzzy (NF), fuzzy logic (FL), ANN-genetic algorithm (GA), GA, multi-objective optimal control (MOOC), expert system (ES), and Bayesian network (BN) are the common AI techniques applied for wastewater treatment, specifically for pollutant removal, reducing the cost and energy consumption during the treatment process, and operation management during wastewater treatment. ANNs can solve multivariate nonlinear problems and are frequently employed in experimental designs to eliminate contaminants during wastewater treatment.
Many ANNs, such as multilayer perceptron (MLP), radial basis function (RBF), feedforward neural network (FNN), wavelet neural network (WNN), self-organizing map (SOM), Elman neural network (ENN), and recurrent neural network (RNN), are utilized for establishing models and wastewater treatment process simulation. Other AI techniques used in wastewater treatment include model tree (MT), clustering algorithm, particle swarm optimization (PSO), support vector machine (SVM), adaptive neuro-fuzzy inference system (ANFIS), and data mining.
For instance, ANN, ANFIS, ANN-GA, MLP-ANN, and FL have been used in aeration, nitrification and denitrification, anoxic-oxic biological, anaerobic digestion, and Fenton oxidation, aeration diffusion, respectively, for conventional pollutant removal. Similarly, RBF-ANN, ANFIS, ANN, MLP-ANN, and backpropagation ANN-PSO can be utilized in membrane treatment, adsorption, phytoremediation, adsorption, and extraction, respectively, for the removal of typical pollutants like copper, cadmium, and lead.
MLP-ANN/RBF-ANN, backpropagation ANN, and stack denoising autoencoder have been employed in sequencing batch reactors, adsorption, and anaerobic-oxic biological treatment, respectively, for mixed pollutant removal. AI technologies like MOOC, data mining/reinforcement learning/ANFIS, ANN, and FL can be used in activated sludge, aerobic biological, wastewater collecting and grid, and continuously stirred tank reactors, respectively, for energy reduction during wastewater treatment.
Similarly, reinforcement learning, NF, rough set-based fuzzy control system, and ES are used in activated sludge, aerated submerged biofilm, sequencing batch reactor, and oxic reactor, respectively, for operational cost reduction during wastewater treatment. Moreover, ANFIS/ANN/MT, NF, data mining, MLP-ANN/RBF-ANN, RBF-ANN/least squares SVM, and SVM/ANFIS/RBF-ANN can be used for several operational management objectives, including aeration efficiency, management of anaerobic and pump systems, sludge bulking, membrane fouling, and daily flow rate management.
AI Applications in Wastewater Treatment
BOD and COD Monitoring: ANN can assess wastewater treatment performance by evaluating variables like total suspended solids, COD, and BOD. BOD determination has been performed for wastewater biological treatment units using multilayer network and link networking analysis methods.
An ANN-based model has been recently developed for the BOD elimination process. An ensemble neural network model has been proven effective for predicting BOD and COD along with total nitrogen in the effluent reliably. Thus, the model can evaluate the wastewater treatment facility's performance.
The application of AI has been investigated for modeling the degradation of penicillin-type antibiotics like cloxacillin, ampicillin, and amoxicillin in aqueous medium. The ANN-generated results displayed a 0.997 correlation coefficient, which confirmed the viability of the model.
Elimination of Substances and Contaminants: A study analyzed the adsorption characteristics of substances like resorcinol and phenol using ANN and compared the results with batch experimental results.
The input variables were the concentrations of the adsorbates, amount of the carbonaceous adsorbent, pH, and time, while the neural network output provided the contaminants’ removal efficiency, which was in corroboration with the experimental results.
AI models like ANFIS and SVM can investigate the nitrogen’s real-time removal efficiency from a wastewater treatment plant. Influent nitrogen, ammonia, COD, total solids, and the pH of the influent were utilized as the input variables. The SVM showed a better forecasting efficiency compared to ANFIS.
Process Parameter Regulation and Treatment Plant Efficiency: ANN and a fuzzy inference system, in conjunction with Kohonen mapping, were utilized to investigate the activated sludge process modeling. Results showed the superiority of AI over conventional methods with higher flexibility and viability.
An AI model containing two phases, including the analytical phase and the synthesis phase, has been developed for wastewater treatment. The analytical phase involved the algorithm extracting conventions from the input and recognizing the treatment methods’ influence at mixed concentrations. The second phase formulated the optimal conditions based on the analytical phase.
AI can be applied to optimize the processing of sewerage systems as AI systems and automation technologies are more economical compared to conventional engineering methods. The AI-based non-linear model enables automatic setpoint rescheduling for a changeable degree of flow conditions in sewage treatment plants. Thus, the filtration efficiency of sewage water can be easily regulated using controller operations, which assists in predicting the future and past behavior of the process.
Water Quality Parameter Determination: ANN/Kohonen self-organizing feature map has been employed successfully for assessing the treatment plants and understanding the process dependencies on variables like concentrations, pH level, and dissolved oxygen. Hardness, turbidity, chloride ion, and pH predictions made using ANN enable the designing of cost-related issues and amounts for wastewater treatment. Models like SVN and ANN are suitable for quantifying the flow discharge significance on the accuracy and precision.
Despite the advantages of using AI techniques in wastewater treatment, several challenges exist with their implementation. For instance, applying AI to address urgent problems in wastewater treatment is difficult. Computer technology skills, knowledge about data and statistical sciences, and resources like sufficient data are essential for the application of AI. Additionally, selecting the wrong problem will lead to failure in the application of AI. Thus, an in-depth understanding of the real AI applications in wastewater treatment is necessary.
In summary, while AI significantly enhances wastewater treatment by improving efficiency, accuracy, and cost-effectiveness, its successful implementation requires overcoming challenges such as technical expertise, adequate data, and precise problem selection.
References and Further Reading
Nguyen, X. C., Nguyen, T. T. H., Tran, Q. B., Bui, X., Ngo, H. H., Nguyen, D. D. (2021). Artificial intelligence for wastewater treatment. Current Developments in Biotechnology and Bioengineering, 587-608. DOI: 10.1016/B978-0-323-99874-1.00008-7, https://www.sciencedirect.com/science/article/pii/B9780323998741000087
Zhao, L., Dai, T., Qiao, Z., Sun, P., Hao, J., Yang, Y. (2020). Application of artificial intelligence to wastewater treatment: A bibliometric analysis and systematic review of technology, economy, management, and wastewater reuse. Process Safety and Environmental Protection, 133, 169-182. DOI: 10.1016/j.psep.2019.11.014, https://www.sciencedirect.com/science/article/abs/pii/S0957582019318403
Kamali, M., Appels, L., Yu, X., Aminabhavi, T. M., Dewil, R. (2021). Artificial intelligence as a sustainable tool in wastewater treatment using membrane bioreactors. Chemical Engineering Journal, 417, 128070. DOI: 10.1016/j.cej.2020.128070, https://www.sciencedirect.com/science/article/abs/pii/S1385894720341875
Malviya, A., Jaspal, D. (2021). Artificial intelligence as an upcoming technology in wastewater treatment: a comprehensive review. Environmental Technology Reviews, 10(1), 177–187. DOI: 10.1080/21622515.2021.1913242, https://www.tandfonline.com/doi/abs/10.1080/21622515.2021.1913242