The accurate forecast of wastewater treatment plant (WWTP) key features can comprehend and predict the plant behavior to support process design and controls, improve system reliability, reduce operational costs, and endorse optimization of overall performances. Deep learning technologies as proven data-driven soft-sensors should be developed for WWTP applications to tackle the process of non-linearity and the dynamic nature of environmental data. This study adopts deep learning-based models as softsensors to forecast WWTP key features, such as influent flow, influent temperature, influent biochemical oxygen demand (BOD), effluent chloride, effluent BOD, and power consumption. We constructed six deep learning models derived from long short-term memory (LSTM) and gated recurrent unit (GRU), namely traditional LSTM and GRU, the exponentially smoothed LSTM, and the adaptive version of LSTM and smoothed LSTM. The employment of a smoothed LSTM technique is expected to reduce the outlier effect and to improve forecasting accuracy. Meanwhile, the usage of adaptive deep models will enhance the capabilities of the LSTM to quickly and accurately follow the trend of future data. We compared the performance of these models with Bi-directional LSTM (BiLSTM) and the seasonal decomposition using local regression. The historical records from a coastal municipal WWTP in Saudi Arabia are used to verify the investigated models' effectiveness. The proposed models provide promising forecasting results but require no assumptions on the data distributions. In terms of efficiency, GRU based models converge faster than LSTM based models. In terms of accuracy, the LSTM soft-sensor shows overall the optimal result for all key features followed by the exponentially-smoothed GRU and LSTM. By contrast, the adaptive models achieved the lowest forecasting performance compared to the other models. These findings will benefit practitioners to achieve data-driven WWTP management.

Wastewater treatment plants (WWTPs) are sustainable solutions to water scarcity. 14 As initial conditions offered to WWTPs, influent conditions (ICs) affect treatment 15 units states, ongoing processes mechanisms, and product qualities. Anomalies in 16 ICs, often raised by abnormal events, need to be monitored and detected promptly 17 to improve system resilience and provide smart environments. This paper pro-18 posed and verified data-driven anomaly detection approaches based on deep learn-19 ing methods and clustering algorithms. Combining both the ability to capture 20 temporal auto-correlation features among multivariate time series from recurrent 21 neural networks (RNNs), and the function to delineate complex distributions from 22 restricted Boltzmann machines (RBM), RNN-RBM models were employed and 23 connected with various classifiers for anomaly detection. The effectiveness of 24 RNN based, RBM based, RNN-RBM based, or standalone individual detectors, 25 including expectation maximization clustering, K-means clustering, mean-shift 26 clustering, one-class support vector machine (OCSVM), spectral clustering, and 27 agglomerative clustering algorithms were evaluated by importing seven years ICs 28 data from a coastal municipal WWTP where more than 150 abnormal events oc-29 curred. Results demonstrated that RNN-RBM-based OCSVM approach outper-30 formed all other scenarios with an area under the curve value up to 0.98, which

Monitoring inflow measurements of water resource recovery facilities (WRRFs) is essential to promptly detect abnormalities and helpful in the decision making of the operators to better optimize, take corrective actions, and maintain downstream processes. In this paper, we introduced a flexible and reliable monitoring soft sensor approach to detect and identify abnormal influent measurements of WRRFs to enhance their efficiency and safety. The proposed data-driven soft sensor approach merges the desirable characteristics of principal component analysis (PCA) with k-nearest neighbor (KNN) scheme. PCA performed effective dimension reduction and revealed interrelationships between inflow measurements, while KNN distances demonstrated superior detection capacity, robustness to underlying data distribution, and efficiency in handling high-dimensional dataset. Furthermore, nonparametric thresholds derived from kernel density estimation further enhanced detection results of PCA-KNN approach when compared with parametric counterparts. Moreover, the radial visualization plot is innovatively employed for fault analysis and diagnosis in combination with PCA and delineated interpretable visualization of anomalies and detector performances. The effectiveness of these soft sensor schemes is evaluated by using real data from a coastal municipal WRRF located in Saudi Arabia. Also, we compared the proposed soft sensor scheme with the conventional PCA-based approaches, including standard prediction error, Hotelling's T 2 , and joint univariate methods. Results demonstrate that this soft sensor-based monitoring approach outperforms conventional PCA-based methods.

To operate wastewater treatment plants (WWTPs) with optimized efficiency, influent conditions (ICs) as initial states of inflow fed to WWTPs were monitored to identify potential anomalies that would trigger adverse events or system crash. To employ voluminous measurements for data-driven decisions, the non-linear, non-Gaussian, non-stationary, auto-correlated, cross-correlated, hetero-skedastic, case-specific nature of multivariate environmental datasets must be considered. This research proposed kernel machine learning models, the kernel principal components analysis based one-class support vector machine (KPCA-OCSVM) with various kernels, to learn anomaly-free training set then classify the testing set. A seven-years multivariate ICs time series was introduced with exploratory analysis performed to reveal temporal behaviors and statistical properties. KPCA with polynomial kernels sufficiently output representative features, based on which OCSVM with Gaussian kernels sensitively and specifically identified anomalies in ICs that were previously omitted by WWTP operators. The proposed kernel algorithms surpassed previous linear PCA-based K-nearest-neighbors models, and improved outcomes with limited increase in computation cost. Without requiring linear, Gaussian, stationary, independent, and homo-skedastic qualities from data, the proposed flexible environmental data science approach could be transferred, rebuilt, and tuned conveniently for ICs from different WWTPs.

Energy consumption is vital to the global costs of wastewater treatment plants (WWTPs). With the increase of installed WWTPs worldwide, the modeling and forecast of their energy consumption have become a critical factor in WWTP design to meet environmental and economic requirements. The accurate and swift energy consumption forecasting soft-sensors are not only supportive to the daily electric and financial budgeting by WWTP practitioners on the micro-scale, but also beneficial to local municipal operation and fundamental to regional environmental impact estimation on the macro-scale. Energy consumption in WWTPs is influenced by different biological and environmental factors, making it complicated and challenging to build soft-sensors. This paper intends to provide short-term forecasting of WWTP energy consumption based on data-driven soft sensors using traditional time-series and deep learning methods. Ten data-driven soft sensors, including the ordinary least square, exponential smoothing state space, local regression, auto-regressive integrated moving average (ARIMA), structural time series model, Bayesian structural time series, non-linear auto-regressive, long short-term memory with and without updates, and gated recurrent units have been investigated and compared for WWTP energy consumption forecasting. Energy consumption time-series data from a membrane bioreactor-based WWTP in the middle east is used to evaluate the performances of the proposed soft-sensors. Results showed that ARIMA achieved slightly improved performances, among others. The employment of adaptive deep learning-based soft sensors is expected to enhance the capabilities of the deep models to quickly and accurately follow the trend of future data.