Data to intelligence: The role of data-driven models in wastewater treatment. (1st May 2023)

• Record Type: Journal Article
• Authors: Bahramian, Majid; Dereli, Recep Kaan; Zhao, Wanqing; Giberti, Matteo; Casey, Eoin
• Is Part Of: Expert systems with applications. Volume 217(2023)
• Publication Date: 2023-05-01
• Subjects: Artificial intelligence -- Machine learning -- Modeling -- Optimization -- Deep learning -- Wastewater treatment; AAD Absolute Average Deviation -- ABC Artificial Bee Colony -- ACO Ant Colony Optimization -- AD Anaerobic Digestion -- AI Artificial Intelligence -- AM Attention Mechanism -- ANFIS Adaptive Neuro Fuzzy Inference system -- ANN Artificial Neural Network -- ARIMA Autoregressive Integrated Moving Average -- ASM2d Activated Sludge Model No. 2 with denitrification -- BOD Biological Oxygen Demand -- BSM1 Benchmark Simulation Model No.1 -- CI Computational Intelligence -- CNN Convolutional Neural Network -- COD Chemical Oxygen Demand -- CSTR Complete Stirred Tank Reactor -- DBN Deep Belief Networks -- DE Differential Evolution -- DNN Deep Neural Network -- DO Dissolved Oxygen -- DT Decision Trees -- EFAST Extended Fourier Amplitude Sensitivity Test -- EQI Effluent Quality Index -- FCM Fuzzy C-means -- FFA Firefly Optimization Algorithm -- FL Fuzzy Logic -- FO Forward Osmosis -- FV Factorial Variance -- GA Genetic Algorithm -- GHG Greenhouse Gases -- GP Genetic Programming -- GRU Gated Recurrent Unit -- GWP Global Warming Potential -- GWO Grey Wolf Optimization -- IA Index of agreement -- IMF Intrinsic Mode Function -- IQR Interquartile Ranging -- KPLS-RVM Kernel Partial Least Squares with Relevance Vector Machine -- LS-SVM Least Square Support Vector Machine -- LSTM Long Short-Term Memory -- MAE Mean Absolute Error -- MAPE Mean Absolute Percentage Error -- MASE Mean Absolute Scaled Error -- MB Methylene Blue -- MBR Membrane Bioreactor -- MCR Mass Content Ration -- MG Malachite Green -- ML Machine Learning -- MLR Multi-Layer Regression -- MLVSS Mixed Liquor Volatile Suspended Solids -- MLANN Multilayer Artificial Neural Network -- MLP-PSO Multilayer Perceptron hybridized with Particle Swarm Optimization -- MOSSO Multi Objective Shark Smell Optimization -- MRE Mass Removal Efficiency -- MSE Mean Square Error -- M5T M5 Tree -- NMSE Normalized Mean Square Error -- NSE Nash-Sutcliffe Error -- OCI Operational Cost Index -- OCRNN Over Complete Recurrent Neural Network -- PBIAS Percent bias -- PCA Principle Component Analysis -- PMOC Polar Mobile Organic Compounds -- P-PO4 Phosphate Phosphorus -- PSO Particle Swarm Optimization -- QBPSA Quantum-Behaved Particle Swarm Algorithms -- R Correlation coefficient -- R2 Coefficient of Determination -- RBFNN Radial Basis Function Neural Network -- RF Random Forest -- RMSE Root Mean Square Error -- RO Reverse Osmosis -- RSM Response Surface Methodology -- SBR Sequencing Batch Reactor -- SC Subtractive Clustering -- SCADA Supervisory Control and Data Acquisition -- SI Scattering Index -- SOM Self-Organizing Map -- SVI Sludge Volume Index -- SVM Support Vector Machine -- SVR Support Vector Regression -- TDS Total Dissolved Solids -- THM Trihalomethane -- TLBO Teaching Learning Based Optimization -- TN Total Nitrogen -- TP Total Phosphorus -- TSK Takagi–Sugeno–Kang -- TSS Total Suspended Solids -- VFA Volatile Fatty Acids -- WWTP Wastewater Treatment Plant; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.119453
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: ANN outnumbered other standalone AI models (single models) applied to WWTPs. Hybrid models were relatively more accurate than the standalone models. Most of hybrid models were classified as CI-metaheuristic models. FL was the most suitable model for the incomplete data sets. Despite recent developments, industrial deployment is lacking. Abstract: Increasing energy efficiency in wastewater treatment plants (WWTPs) is becoming more important. An emerging approach to addressing this issue is to exploit development in data science and modelling. Deployment of sensors to measure various parameters in WWTPs opens greater opportunities for exploiting the wealth of data. Artificial intelligence (AI) is emerging as a solution for automation and digitalization in the wastewater sector. This review aims to comprehensively investigate, summarize and analyze recent developments in AI methods applied to the modelling of WWTPs. The review shows that among the standalone models, Artificial Neural Networks (ANN) was the most popular model followed by, in descending order: Decision Trees (DT), Fuzzy Logic (FL), Genetic algorithm (GA) and Support Vector Machine (SVM). In the case of incomplete data, FL was the most frequently used method as it uses linguistic expert rules to find an approximation for the missing data. Regarding accuracy and precision, hybrid models demonstrated relatively better performance than the standalone ones. Among these models, the Machine Learning (ML)-metaheuristic, which integrates an AI model with a bioinspired optimization method, was the most preferred type as it was used in more than 45% of the hybrid models. Correlation coefficient ( R), Correlation of Determination ( R 2 ) and Root Mean Square Error ( RMSE ) were the frequently used metrics for model performance evaluation. Finally, the review shows that despite recent developments, industrial deployment is still lacking. The industrial application requires close interaction of interested parties, among which research institutes, private sector and public sector play an inevitable role. The future research should focus on mitigating the barriers for more in-depth collaboration of interested parties and finding new paths for more cooperative and harmonized activity of them.