Machine learning algorithm as a sustainable tool for dissolved oxygen prediction: a case study of Feitsui Reservoir, Taiwan

Robust clustering-based hybrid technique enabling reliable reservoir water quality prediction with uncertainty quantification and spatial analysis

Machine learning methodology has recently been considered a smart and reliable way to monitor water quality parameters in aquatic environments like reservoirs and lakes. This study employs both individual and hybrid-based techniques to boost the accuracy of dissolved oxygen (DO) and chlorophyll-a (Chl-a) predictions in the Wadi Dayqah Dam located in Oman. At first, an AAQ-RINKO device (CTD+ sensor) was used to collect water quality parameters from different locations and varying depths in the reservoir. Second, the dataset is segmented into homogeneous clusters based on DO and Chl-a parameters by leveraging an optimized K-means algorithm, facilitating precise estimations. Third, ten sophisticated variational-individual data-driven models, namely generalized regression neural network (GRNN), random forest (RF), gaussian process regression (GPR), decision tree (DT), least-squares boosting (LSB), bayesian ridge (BR), support vector regression (SVR), K-nearest neighbors (KNN), multilayer perceptron (MLP), and group method of data handling (GMDH) are employed to estimate DO and Chl-a concentrations. Fourth, to improve prediction accuracy, bayesian model averaging (BMA), entropy weighted (EW), and a new enhanced clustering-based hybrid technique called Entropy-ORNESS are employed to combine model outputs. The Entropy-ORNESS method incorporates a Genetic Algorithm (GA) to determine optimal weights and then combine them with EW weights. Finally, the inclusion of bootstrapping techniques introduces a stochastic assessment of model uncertainty, resulting in a robust estimator model. In the validation phase, the Entropy-ORNESS technique outperforms the independent models among the three fusion-based methods, yielding R2 values of 0.92 and 0.89 for DO and Chl-a clusters, respectively. The proposed hybrid-based methodology demonstrates reduced uncertainty compared to single data-driven models and two combination frameworks, with uncertainty levels of 0.24% and 1.16% for cluster 1 of DO and cluster 2 of Chl-a parameters. As a highlight point, the spatial analysis of DO and Chl-a concentrations exhibit similar pattern variations across varying depths of the dam according to the comparison of field measurements with the best hybrid technique, in which DO concentration values notably decrease during warmer seasons. These findings collectively underscore the potential of the upgraded weighted-based hybrid approach to provide more accurate estimations of DO and Chl-a concentrations in dynamic aquatic environments.

Multi-step forecasting of dissolved oxygen in River Ganga based on CEEMDAN-AdaBoost-BiLSTM-LSTM model

Accurate prediction of Dissolved Oxygen (DO) is an integral part of water resource management. This study proposes a novel approach combining Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) with AdaBoost and deep learning for multi-step forecasting of DO. CEEMDAN generates Intrinsic Mode Functions (IMFs) with different frequencies, capturing non-linear and non-stationary characteristics of the data. The high-frequency and medium-frequency IMFs, characterized by complex patterns and frequent changes over time, are predicted using Adaboost with Bidirectional Long Short-Term Memory (BiLSTM) as the base estimator. The low-frequency IMFs, characterized by relatively simple patterns, are predicted using standalone Long Short-Term Memory (LSTM). The proposed CEEMDAN-AdaBoost-BiLSTM-LSTM model is tested on data from ten stations of river Ganga. We compare the results with six models without decomposition and four models utilizing decomposition. Experimental results show that using a tailored prediction technique based on each IMF's distinctive features leads to more accurate forecasts. CEEMDAN-AdaBoost-BiLSTM-LSTM outperforms CEEMDAN-BiLSTM with an average improvement of 25.458% for RMSE and 37.390% for MAE. Compared with CEEMDAN-AdaBoost-BiLSTM, an average improvement of 20.779% for RMSE and 28.921% for MAE is observed. Diebold-Mariano test and t-test suggest a statistically significant difference in performance between the proposed and compared models.

Application of machine learning techniques to predict groundwater quality in the Nabogo Basin, Northern Ghana

The main objective of this study was to map the quality of groundwater for domestic use in the Nabogo Basin, a sub-catchment of the White Volta Basin in Ghana, by applying machine learning techniques. The study was conducted by applying the Random Forest (RF) machine learning algorithm to predict groundwater quality, by utilizing factors that influence groundwater occurrence and quality such as Elevation, Topographical Wetness Index (TWI), Slope length (LS), Lithology, Soil type, Normalize Different Vegetation Index (NDVI), Rainfall, Aspect, Slope, Plan Curvature (PLC), Profile Curvature (PRC), Lineament density, Distance to faults, and Drainage density. The groundwater quality of the area was predicted by building a Random Forest model based on computed Arithmetic Water Quality Indices (WQI) (as dependent variable) of existing boreholes, to serve as an indicator of the groundwater quality. The predicted WQI of groundwater in the study area shows that it ranges from 9.51 to 69.99%. This implied that 21.97 %, 74.40 %, and 3.63 % of the study area had respectively the likelihood of excellent. The models were found to perform much better with an RMSE of 23.03 and an R2 value of 0.82. The study conducted highlighted an essential understanding of the groundwater quality in the study area, paving the way for further studies and policy development for groundwater management.

A comprehensive review on artificial intelligence in water treatment for optimization. Clean water now and the future

Given the severe effects that toxic compounds present in wastewater streams have on humans, it is imperative that water and wastewater streams pollution be addressed globally. This review comprehensively examines various water and wastewater treatment methods and water quality management methods based on artificial intelligence (AI). Machine learning (ML) and AI have become a powerful tool for addressing problems in the real world and has gained a lot of interest since it can be used for a wide range of activities. The foundation of ML techniques involves training of a network with collected data, followed by application of learned network to the process simulation and prediction. The creation of ML models for process simulations requires measured data. In order to forecast and simulate chemical and physical processes such chemical reactions, heat transfer, mass transfer, energy, pharmaceutics and separation, a variety of machine-learning algorithms have recently been developed. These models have shown to be more adept at simulating and modeling processes than traditional models. Although AI offers many advantages, a number of disadvantages have kept these methods from being extensively applied in actual water treatment systems. Lack of evidence of application in actual water treatment scenarios, poor repeatability and data availability and selection are a few of the main problems that need to be resolved.

Application of data-driven models to predict the dimensions of flow separation zone

In this research, the effect of a submerged multiple-vane system on the dimensions of flow separation zone (DFSZ) is assessed via 192 measured datasets. The vanes' shape comprised two segments, curved and flat plates which are located in the connection of main channel to the lateral intake channel with an angle of 55°. In this direction, a butterfly's array for the vanes' arrangement along with different main controlling factors such as distances of vanes along the flow (δ_l), degree of curvature (β), and angles of attack to the local primary flow direction (θ) is utilized. Through capturing photos and utilizing AutoCAD and SURFER software, maximum relative length and width are calculated. Based on the experimental measurements, maximum percentage reduction of DFSZ, in comparison with the controlled test (without submerged vanes), is obtained with θ = 30°, β = 34°, and δ_l = 10 cm with value of 78 and 76%, respectively. Moreover, several data-driven models, namely, gene expression programming (GEP), support vector regression (SVR), and a robust hybrid SVR with an ant colony optimization algorithm (ACO) (i.e., hybrid SVR-ACO model), are developed in order to predict DFSZ via the operative dimensionless variables realized by Spearman's rho and Pearson's coefficient processes. In accordance with the statistical metrics, model grading process, scatter plot, and the hybrid SVR(RBF)-ACO model are preferred as the best and most precise model to predict maximum relative length and width with a total grade (TG) of 6.75 and 5.8, respectively. The generated algebraic formula for DFSZ under the optimal scenario of GEP is equated with the corresponding measured ones and the results are within 0-10%.