Enabling wastewater treatment process automation: leveraging innovations in real-time sensing, data analysis, and online controls

Wenjin Zhang *a, Nicholas B. Tooker b and Amy V. Mueller a
aDepartment of Civil and Environmental Engineering, Northeastern University, 360 Huntington Ave., Boston, MA 02115, USA. E-mail: zhang.wenji@northeastern.edu; Fax: +617 373 4419; Tel: +617 373 2444
bDepartment of Civil and Environmental Engineering, University of Massachusetts Amherst, 130 Natural Resources Road, Amherst, MA 01003, USA

Received 26th April 2020 , Accepted 10th August 2020

First published on 7th September 2020


The primary mandate of wastewater treatment facilities is the limitation of pollutant discharges, however both continued tightening of permit limits and unique challenges associated with improving sustainability (i.e., resource recovery) demand innovation. Enabling increasingly sophisticated treatment processes in a cost-effective and energy-efficient way requires expanding capabilities for rapid, accurate real-time quantification of a broadened range of wastewater constituents as well as envisioning novel feedback control strategies based on these signals. This manuscript quantitatively compares results of early adoption of instrumentation and process upgrades at operating wastewater treatment facilities and proof-of-concept research results, with a focus on leveraging real-time sensing of wastewater chemistry for process monitoring and control. Up to 10% improvement in nutrient removal and energy savings are already being achieved, yet shortfalls in hardware readiness, lack of field-relevant context of research results, and a widening gap between the training of environmental engineers and the skillsets required to develop and maintain sensor-driven solutions present challenges. A forward-looking roadmap highlights opportunities for accelerating innovation, including (1) ensuring research results are published in units and context that allow operators to make an informed cost–benefit analysis with explicit comparison to existing operational baselines, (2) promoting integrated design of hardware and software to generate novel approaches for improved sensing of target analytes, (3) strengthening partnerships nationally, including for data sharing, field testing of new hardware, and expanding educational curricula, and (4) building forums for sharing of expertise and data among plant operators to enable smaller facilities to more cost-effectively collect information required to design and evaluate upgrades.


image file: d0ew00394h-p1.tif

Wenjin Zhang

Wenjin Zhang is currently a Ph.D. candidate in Interdisciplinary Engineering in the Department of Civil & Environmental Engineering at the Northeastern University. Her research focuses on informatics for decision and control of civil and environmental infrastructure systems, including machine learning applied to wastewater treatment sensing and process control.

image file: d0ew00394h-p2.tif

Nick Tooker

Nick Tooker is a Professor of Practice in the Department of Civil & Environmental Engineering at the University of Massachusetts Amherst. He has experience in design and research of nutrient removal and recovery systems for water resource recovery facilities. In 2019, he received the Ralph Fuhrman Medal for Outstanding Water Quality Academic-Practice Collaboration.

image file: d0ew00394h-p3.tif

Amy Mueller

Amy Mueller received the S.B. (′02) and M.Eng. (′03) degrees in electrical engineering and computer science and the Ph.D. (′12) degree in environmental chemistry from the Massachusetts Institute of Technology, Cambridge, MA, USA. She is currently an Assistant Professor jointly appointed in the Colleges of Engineering and Science at Northeastern University, Boston, MA, USA. Her primarily field of research is sensor and control system development for study and optimization of biogeochemistry of natural systems and urban infrastructure.



Water impact

Stringent pollutant regulation and a vision of future water resource recovery demand innovation in wastewater treatment. Implementation of novel sensor-driven processes shows 10% improvement in nutrient removal and energy savings, but broadened stakeholder partnerships (e.g., data sharing, field testing of hardware, education design) are critically needed to translate adoption of such processes at a wider range of facilities.

1 Introduction

The primary mandate of wastewater treatment facilities is the limitation of pollutant discharges1 (i.e., primarily nutrients, organic carbon substrates, and metals) in accordance with legislation like the Clean Water Act2 and the European Union Water Framework Directive (WFD).3 Both the continued tightening of permit limits (e.g., the 2018 revision to the National Pollutant Discharge Elimination System (NPDES)4) and the unique challenges associated with improving the sustainability aspects of water treatment (i.e., resource recovery and reuse) are driving innovation in this field.5,6 For instance, novel microbial consortia (e.g., anammox7,8 and phosphate accumulating organisms (PAOs)9) are already providing alternatives, or complementary methods, to the traditional nitrification–denitrification scheme; such strategies can decrease aeration costs (which can represent up to 50% of total plant energy use10,11), decrease use of metals for phosphate removal, and/or increase opportunities for nutrient recovery.12–14 Recently, the potential for wastewater treatment facilities to play a role in moving toward a closed-loop approach on water – and the constituents of wastewater, such as nutrients and energy – is also gaining recognition, with implementation of strategies such as nutrient recovery and direct (or indirect) potable water reuse exemplifying how wastewater treatment plants are being re-envisioned as wastewater resource recovery facilities (WRRF).15

Enabling these increasingly sophisticated treatment processes in a cost-effective and energy-efficient way demands innovation in online, real-time techniques for process monitoring and automated process control – as does the automation necessary for improving cost-effectiveness and sophistication of processes at small or decentralized treatment facilities.16 Achieving these goals requires expanding capabilities for rapid, accurate quantification of wastewater constituents as well as innovating in feedback control strategies based on either traditional or novel sensing modalities,17 while at the same time balancing tradeoffs between complexity and management of potential failure modes.

One strategy for accelerating progress in these efforts is through increased visibility of existing tools and successes. For instance, recent reviews provide a summary of the current state of nutrient monitoring methods (nitrogen and phosphorus)18 as well as insight into the potential for use of biosensors19 and fluorescence spectroscopy20 in wastewater monitoring. Wastewater process control technologies21 and methods22,23 have also been discussed, increasingly promoting the use of big data and statistical process control to address unique characteristics in wastewater treatment.24,25 Different teams have focused on the wide range of options available in leveraging big data, e.g., regression models for biological process variable prediction26,27 or resource recovery,28 plant-wide process modeling and monitoring,29 and development of multiple sensor-driven inference models.30

Largely, however, these efforts have focused primarily on emerging research and the long-range future of the discipline, which has created a gap wherein limited actionable information is available for plant operators to assess which technologies are ready to be adopted, what the outstanding challenges are, and what timelines might be feasible for integrating these innovations into operational strategy. This manuscript seeks to bridge this gap by considering the practitioner needs and identifying imminent roadblocks preventing the flow of innovation into use at operating wastewater treatment facilities, updating similar efforts undertaken as the instrumentation boom began in the early 2000s.31 This is achieved through an integrated analysis of case studies for which quantitative results are reported, both in operating plants and in research that leverages pilot plants or uses operational data from wastewater treatment facilities. The results are evaluated in the context of user needs (focusing primarily on wastewater chemistry of secondary treatment processes, where instrumentation is driving impactful progress toward WRR goals) to identify gaps and recommend strategies for near-term critical research and for better leveraging expertise of all stakeholders, from plant operators and engineers to researchers to commercial instrumentation providers, to collectively work toward future solutions.

Briefly, the manuscript is laid out as follows. Section 2 provides an overview of wastewater treatment processes and introduces relevant terminology as context for the reader. Section 3 describes strategies used and results achieved at operating treatment facilities using sensors and real-time controllers primarily for improved nutrient removal and energy efficiency. Section 4 highlights current and ongoing research on novel sensors, data analytics methods, and control schemes for use in treatment facilities, highlighting studies for which practically interpretable numerical results available. Section 5 proposes targeted areas for continued research to support innovation in process design and control, seeks to understand how to improve transfer of innovations from academic research to operations, and suggests opportunities for progress built around collaborative efforts. Finally Section 6 concludes with a summary of key insights.

2 Wastewater treatment: context and terminology

2.1 Wastewater process overview

Wastewaters can be roughly divided into two categories based on chemical makeup: industrial and municipal.1 Industrial wastewaters are highly variable depending on the source process, can include a wide range of potentially toxic materials (e.g., lead, silver, mercury, copper, nickel, chromium, zinc, cadmium, or tin32), can have widely varying pH (e.g., as a consequence of use of strong acids or bases for cleaning17), and/or can have high biochemical oxygen demand (BOD).33 Because of the variability in treatment needs, industrial discharges are generally mandated to be actively managed on site (e.g., through pre-treatment) before discharge into municipal sewers. In contrast municipal wastewaters, i.e., the collected waste streams of residences, non-industrial workplaces, and pretreated industrial effluents, are typically high in organic content and nutrients which are therefore the primary targets of municipal wastewater facilities. Because of the relative similarity in aims at most municipal plants, the more variable range of industrial treatment purposes, and the relative lack of available publications on industrial treatment processes (which tend to be managed by private commercial entities), the case studies in this manuscript will focus primarily on municipal treatment processes.

In municipal wastewater facilities, the secondary treatment stage is where carbon and nutrient removal are achieved, most frequently by using biological processes.5 Because of the need to maintain a stable microbial population and due to the high cost of aeration (as a fraction of plant operating cost), optimization efforts both in deployment of sensors and in control strategies have focused mostly on secondary treatment, and innovation continues to focus on this process stage as the most impactful area for improvement.

2.2 Terminology

Typically innovations in wastewater biological treatment processes go through one or more development or test phases at smaller scales before being implemented in operating plants. Lab-scale reactors (small, e.g., size of ≤5 liters) may be operated at municipal facilities or within academic labs, often on synthetic wastewater (i.e., stable conditions that allow controlled testing and evaluation during development). Pilot-scale reactors (e.g., hundreds to thousands of liters) are typically operated as a sidestream at municipal facilities to evaluate usability on real wastewater streams. Pilot-scale systems often represent collaborations between municipal, academic, and/or engineering consulting (e.g., sensor or controller design) partners, may represent a scale-up of lab-scale work, and are generally a good test for stability in the face of real wastewater variability. Full-scale generally refers to any process implemented in the main treatment train at a municipal utility, however it is important to note that the size range for such facilities is extremely wide (i.e., from <1 to >100 million gallons per day (MGD) processed waste flow or, alternatively, serving <2000 to hundreds of thousands of people).

Usage of terminology related to sensor and data approaches can vary; this manuscript will follow published references (e.g., ref. 34–36), briefly summarized here. A sensor is a type of instrument that responds to a property of the environment, reporting in real time as an analog or digitally readable signal (e.g., a voltage or ASCII data stream). An analyzer is an instrument that measures some property of a sample after completing some chemical alteration, e.g., mixing with a reagent, and results typically have a time delay (often 5–10 minutes but up to 30 minutes when chemical digestions are required). Real-time refers to data that are available very shortly (typically <1 min) after requesting a reading, while online refers to an approach that digitizes data instantaneously to make it available for post-processing or use (noting this does not necessitate it be web-linked, just that the data be available to a local processing unit or controller). Logged data refers to the data recorded over a period of time by a computer system or by lab staff which can later be analyzed both for retrospective investigations or prospective testing of new approaches. Chemometrics is the umbrella term capturing the application of mathematical, statistical, and computation methods to address questions related to system chemistry. Soft sensor refers to a system that utilizes a tuned algorithm to estimate the concentration of a target analyte from readings from multiple individual sensing devices or, alternatively, to estimate from a single sensor one or more outputs which are not measured directly by the sensor being used.37–39 In contrast with traditional sensors, which (ideally) respond only to and directly to a single property of the environment, soft sensors leverage responsiveness to multiple environmental properties, indirect relationships, and/or expert knowledge embedded in models to improve quality of the estimate of the target analyte. As a consequence the associated algorithms are generally more complex (have more tuning parameters) than a standard calibration curve.

3 Current state of sensing and control strategies in treatment facilities

Many wastewater treatment plants already utilize sensors for pH, dissolved oxygen (DO), oxidation–reduction potential (ORP), and/or electrical conductivity (EC) for monitoring and, in some cases, automated process control,40,41e.g., maintaining DO levels through control of blowers in response to real-time DO sensor readings.42 In the past decade, a major focus of operational interest has been direct measurements of nutrients and integration of these data into control strategies. The following sections provide an overview of current capabilities through a lens of case studies from a number of facilities implementing novel uses of sensors in conjunction with innovation in data processing and controls.

3.1 Innovative sensing approaches in operating wastewater treatment facilities

As expanded options for real-time sensing technologies become increasingly available, many treatment districts are developing pilot studies (some in collaboration with academic researchers) or upgrading operational facilities43 to (1) improve nutrient sensing for process characterization and (2) optimize oxygen delivery (i.e., achieve energy efficiency gains); Table 1 provides an overview of projects discussed in this section.
Table 1 Examples of real-time sensing applied in operating WWTPs. * indicates sensor type not reported. Abbreviations listed in the table follow. CSTR: continuously stirred tank reactor; MLE: modified Ludzack–Ettinger; PFR: plug-flow reactor; SBR: sequencing batch reactor; TSS: total suspended solids
Treatment plant Location Type of process Scale Year sensor-driven strategy implemented Highlighted sensing strategy Reported results
Källby WWTP44 Lund, Sweden Pre-denitrification Full-scale 2001 DO (optical), NH4–N (ISE), nitrate (ISE) 13% energy savings annually
City of Layton, WWTP45 Layton, FL Nitrification/denitrification (SBR) Full-scale 2009 DO*, ORP (electrode), TSS* TSS:
Pre-upgrade: 89.3 mg L−1
Post-upgrade: 64.1 mg L−1
Nitrogen:
Pre-upgrade: 7.88 mg L−1
Post-upgrade: 3.33 mg L−1
Strass WWTP46 Strass im Zillertal, Austria Nitritation/denitritation (SBR) + sidestream anammox for deammonification Full-scale 2011 NH4–N (ISE) Over 30% energy savings
HRSD WWTP47 Virginia Beach, VA Nitritation/denitritation (CSTR sidestream) Pilot-scale 2012 DO (optical), NO2/NO3 (UV/vis), NH4–N (ISE) Acetate savings of 60% compared to conventional nitrification–denitrification
WSHD48 Rotterdam, Netherlands Partial nitritation–anammox (PFR) Pilot-scale 2012–2013 Automated analyzer for NO3 10% energy savings
Wildcat Hill WWTP45 Flagstaff, AZ Nitrification/denitrification (MLE) Full-scale 2013 Combined NO3 and NH4+ (ISE) Nitrogen (TN) effluent:
Pre-upgrade: 14 mg L−1
Post-upgrade: 8.5 mg L−1
Ejby Mølle WWTP49 Odense, Denmark Nitrification/denitrification (CSTR sidestream) Full-scale 2014 DO (optical), NO2, NO3, NH4+ (UV/vis) 18% energy savings, 56% decreased N2O emissions
Municipal WWTP of Manresa50 Barcelona, Spain Nitrification stage of multi-stage process (SBR) Pilot-scale 2017 DO (optical), combined NO3 and NH4+ (ISE) 25% reduced operating costs


In particular, ammonium and nitrate sensors (ion selective electrodes (ISEs)) marketed for use in wastewaters have greatly improved in recent years as companies have created multi-ISE soft sensor packages that improve accuracy of nutrient measurements by internally accounting for sensor interferences (e.g., ref. 51–53). These devices have already been leveraged for a range of purposes in wastewater process control. For instance, nitrate sensors have been used to (1) stabilize nitrate concentrations in the anoxic zone of a nitrification–denitrification process (municipal WWTP in Sluisjesdijk, Rotterdam48 and Ejby Mølle WWTP in Denmark49), (2) monitor sludge recycle (City of Layton WWTP, Florida45), and (3) allow continuous monitoring of both primary and secondary effluents (Wildcat Hill Wastewater Treatment Plant in Flagstaff, Arizona45). Ammonia sensors have been used to (1) inform DO controller rates in a pilot-scale deammonification process (Blue Plains Advanced Wastewater Treatment Plant (AWTP) in Washington, DC54) and a pilot-scale nitritation process in Municipal WWTP of Manresa, Spain,55 (2) optimize anammox processes (Hampton Roads Sanitation District in Surry, Virginia47,56,57), and (3) underpin ammonia-based aeration control (ABAC) (e.g., at Hampton Roads58 and Inland Empire Utilities Agency in Chino, California,59,60 and Källby WWTP, Sweden44). In the case of ABAC, an aeration strategy attracting a great deal of attention recently,61 major cost savings have so far been demonstrated. Hampton Roads reported an overall decrease of 10% in energy consumption, of 53% in supplemental carbon demand for denitrification, and of 40% reduction of final effluent ammonia (monthly aggregated quantities),58 while Inland Empire reported achieving comparable energy savings (∼10%).60 Innovation beyond ABAC has been considered by using real-time ammonia, nitrate, and nitrite sensors together to control aeration in an ammonia verses nitrate (AVN) configuration, with preliminary results suggesting improvements possible relative to ABAC (e.g., 8.6% reduction in oxygen demand and 17.5% increase in nitrogen removal62).

In contrast, reported efforts related to phosphorous removal have primarily taken advantage of high resolution data from online analyzers to inform manual process parameter adjustment rather than real-time controls, e.g., at the Upper Blackstone Water Pollution Abatement District in Worcester, Massachusetts63 and the Westfield (MA) Water Recovery Facility (WRF).64

3.2 Data processing and controls in WWTPs

To ingest and respond to data streaming from commercial sensing technologies, treatment plants are adopting increasingly sophisticated automated real-time data analysis and advanced feedback control systems. In some cases, complex data analytics such as machine learning approaches are being tested to help wastewater treatment facilities improve decision making and online process characterization,65,66 however adoption remains at an early stage and/or proprietary to the implementing companies, making its adoption poorly represented in the published literature.

In contrast, novel and tested control schemes have been more widely published for this application. Due to the highly nonlinear nature of wastewater treatment processes and component responses (e.g., butterfly valves, pumps, biological metabolisms), advanced controllers are in many cases better suited to accounting for component characteristics, time lags, or hysteresis in system response than simple set-point controls.67 Such methods, like PID (proportional integral derivative) control which has been widely used in industrial processes,68 including wastewater treatment plant applications, for decades,69,70 smooth performance by taking into account the magnitude of deviation from the target value and projected rates of change, and can be especially effective when coupled with upgraded hardware such as variable frequency drives (VFD) for pump speed control.

For treatment process control this has been particularly successful for the stabilization of dissolved oxygen (DO) in secondary treatment bioreactors71 and overall management of complex processes (e.g., activated sludge plants) by using multiple PID control loops (e.g., for independent control of recycle pumps, DO blowers, etc.).72 Ammonia-based aeration control (ABAC), introduced above, uses in situ ammonia measurements to calculate deviation from a target setpoint, and a PID controller that takes into account the process dynamics (i.e., O2 required to achieve desired change in ammonia) is typically implemented to achieve the setpoint.

3.3 Lessons learned and remaining gaps

The above studies provide some insight for operators in expanding instrumentation beyond well-established approaches leveraging vetted technologies such as pH, DO, and ORP probes. In some cases online sensors supplement measurements otherwise done in the lab, e.g., TSS sensors can provide a surrogate for laboratory mixed liquor suspended solids (MLSS) analysis. The major success of previous decades, however, is leveraging sensors for ammonium and nitrate to achieve more consistent effluent total nitrogen (TN) concentrations and significantly decreased oxygen consumption. Well-detailed examples of approaches particularly for both mainstream nitrification/denitrification and sidestream anammox are available as a starting point to roadmap potential improvements in other facilities, supported by significant improvements to commercially-available options for sensing nitrate and ammonium from multiple different companies.
Gaps. Despite these successes, however, many operating plants – particularly smaller facilities – still rely on manual measurements of process chemistry and efficiency to inform manual operator control/updating of pump and blower settings, particularly in cases where sensors have failed to meet reliability requirement or exceed available operational budgets.10 Given the heavy human resource overhead, need for toxic chemical reagents, and long time delays associated with lab analysis of organic carbon (minutes for total organic carbon (TOC), hours for chemical oxygen demand (COD), and days for BOD),73–76 there remains great interest in use of organic carbons sensors both for process monitoring (e.g., recovery of BOD for sustainability improvement,77 COD for efficient process control,78 TOC for monitoring impending breakthroughs and upsets/recoveries79) and for compliance with regulatory reporting requirements, yet deployment of emerging commercial options in operating plants remains limited (further details provided below). The lack of commercial sensors for phosphorous despite ongoing efforts (e.g., ref. 80 and 81) is a major gap, with achievable resolution from instrumentation (i.e., ∼30 minutes for total phosphorous (TP)) insufficient to support process control.
Looking forward. As regulations continue to evolve, another potential need in coming years is detecting metals (many of which are already covered by existing regulations) and emerging contaminants in wastewater, e.g., copper, lead, nickel, zinc, mercury, antibiotics, pharmaceuticals, and personal care products.82 In addition, some forward-looking facilities are starting to raise concerns regarding the contribution of treatment processes to global warming, e.g., as N2O emission have been correlated with process parameters such as BOD and NH3–N load.83,84 Preliminary survey of 12 plants across the US over a two year period has suggested ways to meet discharge regulations while minimizing N2O emission,85 and therefore measurements that support monitoring of N2O as well as enabling these processes may be of upcoming interest.

4 Research in sensing and control strategies for WWTPs

Reviews published in the early 2000s suggested the major focus of research to support application of online sensing to wastewater treatment should be (1) adaptation for working in the complex wastewater environment and (2) improving ability to monitor organic matter.86,87 Although some progress has been made on these points, TOC, COD, and BOD remain prime targets for sensor research, with goals of decreasing costs and improving generalized application (reducing need for site-by-site calibrations). In addition phosphorous has been added as a major parameter of interest for online monitoring in the past decades.81,88,89

Other motivations for research on sensors, including those in competition with existing commercial products, include improving accuracy, improving selectivity, decreasing upfront costs, increasing service lifetime, and decreasing maintenance requirements (and therefore associated O&M costs). In particular, both purchase and recurring costs remain major constraints, relative to capital improvement and payroll budgets, resulting in mainly larger treatment facilities at the forefront of adding or upgrading online sensing systems.45 Therefore, decreases in any sensor-related costs may provide critical opportunities for upgrading treatment processes at smaller facilities.

In this context the following sections present emerging research on both hardware and software solutions that represent potential avenues for overcoming challenges currently faced in the design and implementation of novel treatment processes, highlighting options that may be usable in the relatively near future.

4.1 Research in sensing technologies

Active research in sensor technologies applied specifically in the wastewater context includes novel approaches to real-time measurement of a number of parameters, including nitrate/nitrite and ammonia/ammonium, TOC/COD/BOD, volatile fatty acids (VFAs), emerging contaminants, and metals, among others. Promising options already being demonstrated in the wastewater context fall broadly into three categories: optical sensors, ion selective electrodes (ISEs), and biosensors (see Table 2).
Table 2 Sensor types being explored for expanding capabilities in online sensing for wastewater treatment process control and monitoring. Note that advantages, limitations, and highlighted research apply generally to the sensor type
Sensor type Target parameters Advantages Limitations Highlighted research
Optical sensors DO Durability, stability Variable (some high) (Saif et al., 2019)99
Nitrate and nitrite Low maintenance Upfront cost (Shenashen et al., 2013)100
BOD, COD Fast response Variable maintenance costs (Sivchenko et al., 2018)101
Metals
Images (e.g., flocs)
ISEs Nitrate, ammonium Fast response Poor specificity (Bakker, 2019)103
Metals Low upfront cost Sensor drift (Kim et al., 2019)109
Low power demand Limited lifetime (Elbalkiny et al., 2019)108
Biosensors BOD, COD Low upfront cost Short lifetime (Adekunle et al., 2019)118
Metals High selectivity Poor stability (Zhao et al., 2018)120
High maintenance costs (Saberi et al., 2019)126


Optical sensors. Optical sensors based upon principles of absorbance or fluorescence at one or more wavelengths, are typically relatively durable and require minimal maintenance (aside from cleaning); this has resulted in optical DO sensors becoming commonplace in process monitoring and exploration of these principles for other sensing purposes. In particular, optical sensors are a promising option for the measurement of dissolved organic matter,90 an approach studied since the 1990's91,92 but recently becoming more cost-effective and accurate as prices of electronics have dropped and increasingly complex calibrations utilizing multiple wavelengths have been realized.93,94 Real-time sensors for BOD and COD based on spectral absorbance are already commercially available (e.g., ref. 95–97) and have some proven successes, e.g., a sensor leveraging only two wavelengths achieved COD prediction (R2 = 0.9407) in wastewater from a dairy plant in Minnesota.98

Current research on fluorescence-based optical sensors focuses primarily on expanding the number of properties they can measure, e.g., to include metals included in most permit limits. Preliminary successes have been achieved, such as (1) detection of chromium ion down to 3.8 × 10−9 mol L−1 in real wastewater samples (using a nano-phosphor (EBZO) sensor)99 and (2) detection of multiple metals (Cd2+, Hg2+, Cu2+ and Co2+) down to 49.8–50.5 ppb with 60 second response time (using a silica nanoparticle-based sensor),100 however as yet these innovations are not yet commercialized. As electronics prices have dropped, novel uses of cameras for process monitoring are also being explored. For instance an online sensor prototype based on textural image analysis has already been demonstrated for floc monitoring (RMSE = 0.297 mL s−1 and R2 = 0.76), informing dose calculations for a coagulation process in Skiphelle WWTP (Norway).101

Ion selective electrodes (ISEs). Ion selective electrodes (ISEs) are potentiometric sensors that use selective membranes to detect charged species in water samples. ISE technology is promising due to the fast response times (<1 minute), relatively low equipment cost ($100–$1000), low power consumption (mW), and wide measurement concentration range (5 or more orders of magnitude in linear range).102 In contrast, the poor selectivity of most membranes is problematic in wastewaters, and signal drift may require frequent recalibration.103,104 Commercial products using ISEs for nutrient measurement overcome selectivity issues by designing these probes as soft sensors, explicitly measuring and correcting for interfering analytes onboard before reporting data to the user. Research continues to hone this approach to enable more sophisticated methods to correct for interferences even for minority constituents in wastewaters. e.g., for detection of ammonium at concentrations of 10s of micromolar,105 to provide early warning systems for pollutants/hazards (e.g., diesel fuels, antibiotics such as sulfonamides and tetracycline, pharmaceuticals such as ibuprofen and carbamazepine) at the inlet of sewage systems.106–108 It is notable here that these sensor advancements achieved are primarily due to advanced data analytics approaches used (see section 4.2) rather than innovation in the materials or hardware of the sensors. Innovations on that front are also being pursued, e.g., to improve selectivity, reduce equilibration time, minimize sensor drift, extend service lifetime, expand the range of analytes that can be detected.109–111 For example, 30% shorter equilibration time was achieved with novel ion-selective polymer-coated carbon electrodes, demonstrated in a pilot-scale wastewater reactor where these online signals enabled up to a 20% higher NO3 removal efficiency in the final effluent.109 ISEs have also been demonstrated for monitoring of diclofenac, an emerging contaminant, at concentrations down to 0.1 μM and with less interference (∼50% lower selectivity coefficients) than traditional membrane-based ISEs.108 These novel electrode types are in general not, however, in the commercialization process and may therefore benefit from more testing and validation in conditions representative of wastewater treatment processes.
Biosensors. Biosensors leverage living (micro)organisms or biological molecules to detect the presence of chemicals. Potential benefits of biosensors are sensitivity, specificity, low upfront cost, and fast response time (e.g., enzyme-based biosensor response time can be 3–30 seconds (ref. 112)). In wastewater treatment plants, biosensors have been studied to measure specific characteristics of activated sludge processes (e.g., biological materials such as nitrogenous matter113,114 and electrical current produced during the oxidation of hydrogen peroxide115). Currently only very few biosensors are commercially available, specifically for organic matter monitoring in wastewater treatment,116 however published reports of deployments in operating plants are lacking. Despite the limited commercial successes to date, the potential for fast response times and high specificity compared to other technologies on the market continues to encourage active biosensor research, particularly for organic targets such as BOD and COD. A key challenge is expanding the detection range for these analytes.117 Real-time measurement of BOD (range of 0–250 mg L−1, R2 > 0.9) in a wastewater aeration process has been demonstrated using a bioelectrochemical open-type in situ biosensor, while real-time monitoring of COD (range 100–500 mg L−1, R2 = 0.97) in brewery wastewater was achieved using a biosensor based on a microbial fuel cell.118 Research is also expanding the number of targets that can be detected using biosensor sensors, including for (1) nitrogen compounds (e.g., monitoring of NO2via a biosensor based on bacterial reduction of NO2 to N2O, achieving detection limit 1 μM in wastewater with response time 0.5–3 minutes (ref. 119)) and (2) metals (e.g., Cr(VI) monitoring using a biosensor based on a sediment microbial fuel cell, achieving R2 = 0.9935 for concentration range 0.2–0.7 mg L−1 and showing no interference from Cu2+, Zn2+, or Pb2+ in industrial wastewater treatment albeit with ∼18 min response time120), and (3) emerging contaminants (pharmaceutical compound diclofenac measured over range (10–50 mM) in wastewater in real time with a genetically modified yeast-populated foil-based microfluidic flow cell sensor121). Despite this progress, limitations of biosensors can include responses to light,122 a need for high electrical conductivity in sample waters,118 high maintenance needs,123 and the need for a large surface area to achieve usable response magnitudes.115 The need for corrosion resistance, especially in wastewater applications, is also a practical driver of ongoing research.124
Looking forward. For widespread adoption in wastewater treatment processes, sensors should ideally accurately quantify important chemical targets and be user-friendly, durable, accurate and cost-efficient, though in some cases sensors may still be leveraged without achieving all of these characteristics (e.g., if low drift can be guaranteed, accuracy may not be needed to identify changes between operational regimes). For spectrometer-type optical sensors the main issue remains cost, either of the instrumentation itself or the site-specific calibrations required; innovations leveraging only a few wavelengths or low-cost cameras are, however, pushing the envelope and have potential for practical use in the very near term if combined in a scalable way with data analytics (next section). ISE type sensors are already being leveraged in treatment plants, however these also typically require site-specific calibrations and detection limits preclude application to all types of processes or analytes of interest. In this case, in contrast to optical sensors, the inexpensive nature of individual sensor membranes suggests systems with more sensors (also combined with advanced data analytics) have the potential to vastly expand use in wastewaters within the next few years. Relatively speaking, biosensors, while providing potential for a much wider range of targets, remain farther from commercialization and deployment, given outstanding issues such as short reported lifetimes and poor long-term stability.125

4.2 Research in data analytics and controls

In biological treatment processes, microbially mediated reactions coupled with environmental interactions highly nonlinear dynamic systems where traditional modeling approaches may be reaching limits of tradeoffs in accuracy vs. complexity or in ability to accurately calibrate the many parameters. In response a growing number of studies are applying machine learning and statistical methods to wastewater treatment measurement, monitoring, and automation using groups of sensors.34,127,128 Such arrays of sensors, in conjunction with multivariate data analysis (i.e., soft sensors) form a powerful and promising approach for obtaining both quantitative and qualitative information from otherwise complex or cross-interfering data sources,129 particularly given that this approach has already been demonstrated for operationalizing ISE sensors despite outstanding issues of poor selectivity. The most critical challenges in the soft sensor approach are (1) appropriately pairing hardware and software into integrated solutions, (2) identifying data from which to generate calibrations, and (3) designing an accurate method for extraction of useful information from disparate data streams (termed data fusion) – all of which is ideally invisible to the end user who interacts with the system as they would for any out-of-the-box sensor.

Approaches to data fusion that provide an alternative to mechanistic models must be calibrated (“trained”) using data collected to characterize the target application (e.g., biological process),80,113 and therefore transferability of results between sites/applications must always be assessed. However by learning from system data this approach can offer a unique ability to extract information about patterns that may have been previously unidentified or on factors for which no sensor is specifically available.130Table 3 and the following sections highlight several popular data methods as applied to treatment processes to help meet tightening discharge limits131 and to enable real-time provision of a greater number of process parameters.132 While many other applications of these methods exist, the subset selected for discussion were prioritized based on (1) publication in peer reviewed literature and (2) the relative completeness of results (i.e., reporting of coefficient of determination (R2), root mean square, mean square, or mean percent error (RMSE, MSE, MPE), and a comparison with baseline status quo approaches). It is important to note that, while these studies are intended for and in most cases validated with application to wastewater treatment processes, many utilize data which are not yet available via real-time sensing and therefore highlight gains that may be achieved as improved sensor options become available commercially.

Table 3 Recent studies applying machine learning to wastewater treatment challenges, with reported error/predictability. Improvement is defined as the quantitatively improved R2 or decreased RMSE compared to the base case (without the applied model). Definitions of parameters and abbreviations listed in the table follow. ARX: auto-regressive representations with exogenous inputs; CRB: carbamazepine; dose: coagulant dosage; E2: 17β-estradiol; EE2: ethinylestradiol; EF-flow: gaseous emission factor flow; HRT: hydraulic retention time; IBU ibuprofen; MLR: multiple linear regression; morphological parameters: total filamentous bacteria number, filamentous bacteria mean length, total filamentous bacteria length; OLR: organic loading rate; PHI: inlet pH; PHO: pH after coagulant dosage; Qgas: gas flow; QIN: inlet wastewater flow; SVI: sludge volume index; SMX: sulfamethoxazole; TUI: inlet turbidity; TMP: wastewater temperature; TUO: outlet turbidity
Study Sensor type Input Model Output [range] RMSE (unit) R 2 Improvement
(Amaral et al., 2005)133 None Morphological parameters PLS TSS [0.5–4.5 g dm−3] Not reported 0.906 Not reported
SVI [200–620 cm3 g−1] 0.885
(Quintelas et al., 2019)134 Near infrared spectroscopy (NIR) Wavelengths values of IBU, SMX, E2, EE2, CRB PLS IBU [0.05–1.99 mg L−1] 9.17% 0.943 Not reported
SMX [0.31–7.60 mg L−1] 10.20% 0.923
E2 [0–2.75 mg L−1] 10.72% 0.951
EE2 [0.44–6.93 mg L−1] 17.50% 0.858
CRB [0.56–8.12 mg L−1] 8.79% 0.963
(Sivchenko et al., 2018)101 Floc sensor Signals (QIN, PHI, TUI, TMP, dose, PHO, TUO), textural feature of flocs image PLS Coagulant dosages [2.4–7.3 mL s−1] 0.297 mg L−1 0.78 Not reported
(Qin et al., 2012)135 UV/vis spectrometer and a turbidimeter COD, TSS, O&G PLS COD [0–2500 m L−1] 141 mg L−1 0.945 Not reported
TSS [0–450 m L−1] 30.2 mg L−1 0.965
O&G [0–550 m L−1] 34 mg L−1 0.945
(Lee et al., 2008)136 Multi-sensor system Temperature, ORP, pH, DO, flow rate, turbidity, conductivity PCA + ARX + ANN TN [normalized 0–1] Not reported 0.952 RMSE decreased 75% with ARX for COD [0–40 mg L−1]
TP [normalized to 0–1] 0.934
TCOD [normalized to 0–1] 0.921
(Liu et al., 2019)137 pH, ORP, temperature sensor, gas flowmeter pH, ORP, Ogas, CO2, CH4, H2 PCA + LSSVM COD [0–500 mg L−1] 3.08 mg L−1 0.975 Not reported
(Yoo et al., 2003)138 Not reported Flow rate, cyanide, CODin, DO, T, pH PCA + fuzzy regression CODremoved [mean = 605.4 mg L−1] 0.5898 mg L−1 0.9 RMSE decreased 12.85%
(Zounemat-Kermani et al., 2019)139 Odalog logger L2 instrument Q, T, TS, pH, BOD PCA + ANN EF-flow [0–500 mg m−3] 15.26 mg m−3 0.95 RMSE decreased 50% compare to MLR
(da Silva Ribeiro et al., 2019)140 Monopolar electrodes pH, treatment time, current density ANN Boron removal [0–100%] Not reported 0.973 Not reported
(Antwi et al., 2018)141 Spectrophotometer for NH4+–N Influent COD, pH, NH4+–N wavelength, VFA, OLR, biogas yield ANN COD removal [0–100%] 0.104% 0.83 Not reported
(Zaghloul et al., 2018)142 HACH DR for COD, laser particle size analysis system Influent NH4+–N, PO43−, pH, OLR, superficial air velocity, TMP, HRT, setting time ANN COD [0–5000 mg L−1] 0.5198 mg L−1 0.9998 R 2 increased 11%, RMSE decreased 85.7% compare to (Gong, 2018)143 for COD removal prediction
NH4+–N [0–150 mg L−1] 0.2828 mg L−1 0.9987
PO43− [0–60 mg L−1] 0.0383 mg L−1 0.9987


Partial least squares (PLS) regression. Partial least squares (PLS) regression is the most common linear calibration approach in chemometrics,144 used to realign information contained in signals (e.g., sensor signals) with a target set of axes (e.g., chemical concentrations). PLS is particularly useful when traditional one-to-one calibrations do not apply, e.g., when sensors have known interferences or where multi-wavelength or multi-pixel data are collected. Nonlinear PLS is more frequently used in wastewater applications, to capture process changes and dynamics in addition to dealing with mixed sensor signals.145,146 To date this approach has been successfully used to estimate (1) TSS from microscope images of aeration basin samples (R2 = 0.906),133 (2) coagulant dosage in wastewater using a camera-based optical floc sensor prototype (RMSE = 0.297 mg s−1 and R2 = 0.76),101 (3) COD, TSS, and oil & grease (O&G) concentrations in samples of restaurant wastewater from UV/vis spectrometry (R2 of 0.872, 0.928 and 0.931 respectively),135 and (4) pharmaceutical concentrations (e.g., ibuprofen, sulfamethoxazole, 17β-estradiol and carbamazepine) in wastewaters from FT-NIR spectroscopy (R2 = 0.95).134
Principal component analysis (PCA). Principal component analysis (PCA) reduces dimensionality of data by extracting a small number of explanatory factors that describe the major trends in the original data set,147,148 making it a powerful tool for data compression and information extraction in applications that may have many sensor signals.149–151 However the extracted information is not necessarily aligned with target axes, and therefore PCA is most often used as a pre-processing stage for other numerical methods that may require or simply perform better on uncorrelated inputs. The use of PCA in two-stage approaches is a common practice in wastewater process monitoring research, where sensor inputs are often expected to be correlated in some way, and provides another approach for reducing impacts of interferences or complexity of multi-wavelength sensor outputs. For example, real-time wastewater effluent COD, nitrogen, and phosphorus concentrations (R2 of 0.921, 0.952, and 0.934 respectively) were estimated at a small plant in Gyeongbuk, Korea from a 6-sensor suite (pH, DO, ORP, temperature, turbidity, conductivity) by combining PCA (extracting 4 uncorrelated components explaining 82.5% of variability from the 6 interrelated sensor signals) with artificial neural networks (ANN).136 Similarly, COD was estimated (RMSE = 3.08 mg L−1 over range 0–500 mg L−1, R2 = 0.9752) in a lab-scale aerobic wastewater treatment system from 6 data inputs (online monitored gas flow, pH, ORP, H2, CH4, and CO2) by combining PCA (identified 3 components explaining 92.34% of variability) with a least square support vector machine (LSSVM) model.137 A different set of 6 data parameters (sensors for temperature, flow rate, and pH and lab measurements of TSS, BOD, and total dissolved sulfides) was used to estimate the gaseous emission factor flow (EF-flow) for four different WWTPs operated by one collection system in Louisiana by combining PCA and NARX (nonlinear autoregressive with exogenous inputs neural networks), achieving average RMSE = 15.26 mg m−3 (range 4.95–34.11 mg m−3 for 4 plants) and R2 = 0.96 (range 0.962–0.993 for 4 plants).139 Finally, 10% error reduction in estimation of COD removal (CODin–CODout) was achieved in a full-scale plant from 11 inputs (including DO, pH, measured volatile suspended solids, temperature, COD, cyanide, and flow rate from two locations, etc.) by combining PCA with fuzzy regression138 Although less commonly used in wastewater processes, PCA can also be leveraged to better understand complex process correlations, with or without integration with other algorithms. One good example of this is use of PCA with parallel factor analysis (PFA) to characterize cyclic changes in a multistage biological lagoon treatment system in Melbourne, Australia.152
Artificial neural networks (ANNs). Artificial neural networks (ANNs) are powerful tools for complex system modeling and pattern recognition in various fields,153 capable of representing multivariate relationships in highly nonlinear and dynamic systems without requiring any explicit formulation of the mathematical model of the underlying physics. ANNs have already been explored for modeling bioprocesses154,155 and wastewater treatment plants,156 though to date results in the literature represent primarily research studies rather than evaluation for plant operations. In research applications, this approach is pushing the envelope for measuring parameters of interest identified at the start of section 4. For instance, effluent characterization (COD, NH3–N and PO43−) for a lab-scaled aerobic granular reactor processing synthetic wastewater was conducted using an ANN based on 8 input variables (pH, temperature, air velocity, influent NH3–N and PO43−, organic loading rate, HRT and settling time of the sequencing batch reactor); this system achieved RMSE < 5% and R2 > 0.99 for all parameters.142 Finally, ANNs have been used to estimate COD removal from real wastewater samples from an upflow anaerobic sludge blanket reactor using 6 inputs (pH, spectrophotometer measured NH4+, and lab measured COD, effluent VFA, organic loading rate (OLR), and biogas), achieving good agreement with lab measurements (R2 = 0.87 for COD range 42.18–97.93 mg L−1).142
Other approaches. Although the above represent common approaches, a number of other data analytics techniques have also been applied to characterization of wastewater treatment processes in recent years. This includes random forests (e.g., for hourly influent flow rate prediction from historical weather information and flow data for a wastewater treatment plant in Canada, achieving R2 = 0.925 by using time-lag information157) and topological analysis, a variety of graph theory that can be used on large datasets (e.g., for continuous real-time measurement of ammonium and nitrate at the outlet of an aeration plant based on 23 ISE sensors158). Fuzzy logic applications are also being expanded, both for sensing and for controller design (next section). For process characterization, fuzzy logic simulation was evaluated to improve the sequential treatment of wastewater collected from a paper and pulp industry (Rawalpindi, Pakistan) by predicting 4 outputs (total dissolved solids (TDS), TSS, COD, and BOD) from 7 inputs (treatment time, pH, EC, TDS, TSS, BOD, and COD before treatment), achieving R2 of 0.94, 0.89, 0.86, 0.85 respectively.159
Looking forward. The opportunity to expand sensing capabilities in wastewater treatment processes by leveraging advanced data analytics is already clear, with numerous examples above showing application in operating plants and a pipeline of opportunities to replicate those successes as novel hardware becomes commercially available. The essential shared factor across all of these methods is data availability, i.e., this approach relies on historical process data, and therefore the representativeness and completeness of those datasets are critical to ensuring that the model will produce accurate results across all (or a sufficiently large fraction of) possible operational modes or conditions of the facility. In many cases, multiple algorithmic approaches could be applied to the same dataset, but some advantages of each can be identified. The main strength of PCA is as a pre-processing tool to identify (and remove) correlations between signals, to identify relationships between sensor signals and other available information (e.g., lab measurements), or to flag outliers (e.g., extreme events); conversely PCA has little utility if the sensor signals are known to be relatively independent. Both NL-PLS (nonlinear-partial least square) and ANN can be used to approximate complex systems without a need to define underlying physics or controlling parameters; of these two, PLS can provide more specific insight on a mathematical relationship between predictors (i.e., sensors) and outputs (e.g., nutrient concentration) while the structure of an ANN model is less directly interpretable but therein potentially more flexible in what relationships can be represented. ANN models can be particularly useful for large datasets, especially for multi-dimensional data such as images (e.g., photographs of flocs) and non-numerical (categorical) data (e.g., certain weather conditions). A major shortcoming of most academic publications is the lack of comparability between results achieved using novel models and a traditional counterpart (or results achieved using competing model architectures like PLS and ANN); this makes a priori recommendations for model selection in the design of a new project more difficult. In general, it should be noted that the effort required for model development (software creation) is greatly decreased compared to design and programming of a more traditional physics and process-based model, however the need for collection of training data (in some cases up to one or more years of data, if weather conditions vary greatly) may offset this benefit.
Controllers. Innovation in controller algorithms is another promising tool for stabilizing novel plant process architectures. Here fuzzy logic is also highlighted as a method for modeling the complexity and residual uncertainty of biological wastewater treatment processes.160 Many potential gains are demonstrated by evaluating algorithms on logged data from real plants. These include a projected 10% decreased in energy usage (by using fuzzy logic compared to setpoint control) for aeration control based on influent flow rate, COD, and nitrogen at Taradell Wastewater Treatment Plant in Barcelona, Spain,160 a projected increased TN removal efficiency (from 83.7% to 92.5%) and up to 10% energy saving for the biological nutrient removal (BNR) process at Chelas and Castelo Branco WWTP in Portugal,161 and a projected increased phosphorous recovery (29.6% compared to 13.7%) with an elutriation process controlled by fuzzy logic in a lab-scale reactor fed with the thickener supernatant from Calahorra WWTP in La Rioja, Spain.162 In contrast, some successes have already been demonstrated. For instance, a Phaseplit European research program already succeeded in fully automating an entire pilot wastewater treatment facility using fuzzy logic rules, finding that fuzzy logic was more suitable than PID to deal with the variability in the input waste stream and the long setting times in the anaerobic process.163 This required, however, increasing the complexity of the control mechanisms, which was achieved through significant collaboration between plant and research staff – an excellent example of how to successfully approach development of next-generation solutions. More traditional controller architectures (e.g., feed-forward/feedback) are also being leveraged to enable novel processes like ammonia-based aeration control in operating plants and can serve as excellent models for collaborative design processes.164
Looking forward. Application of new techniques for data analytics and online control to wastewater treatment process is already producing real opportunities for energy savings and improved effluent water quality. In particular machine learning algorithms (ANNs, among others) show promise in improving predictability through learning patterns from historical data. However, as one of the characteristics of machine learning algorithms is a requirement for large volumes of representative training data, most studies reported have developed and tested algorithms only on simulated, lab-measured, or (at best) WWTP-logged data. In many of these studies, data used as inputs to the algorithms may not actually be available in real (or near-real) time, making it difficult to envision application in an operating facility. Furthermore publications rarely include a sensitivity analysis that provides insight into the value of each input signal, which would be required to do a cost/benefit analysis for design of required sensors/instruments. A similar issue arises in use of multi-stage algorithms, e.g., cases where PCA is applied as a pre-processing step; because results are rarely provided for models with and without the pre-processing stage, it is difficult to evaluate the trade-offs between complexity and accuracy. On the whole, however, results of preliminary studies are promising and ready for on-site testing. A critical next step in these research arenas will be for researchers to develop more collaborations with operating facilities to evaluate the robustness of these approaches when applied to real-time streaming data in operating facilities with variable inputs and complex dynamics.

5 Discussion

The lack of suitable and robust online sensors for key variables at appropriate detection limits and temporal resolution (e.g., for nutrients) is a major roadblock to the implementation of more advanced bioprocesses and more energy-efficient control schemes. At the same time, there is an increasing desire for high resolution data describing parameters that characterize plant efficacy or are covered in permits (e.g., BOD, COD, N2O, etc.). Further despite progress reported above, plant operators (consulted through a series of stakeholder workshops run by the authors over 2019–2020 and at instrumentation sessions at regional conferences) continue to identify cost (both capital and maintenance) of instrumentation as a significant factor in design and implementation of treatment systems, particularly for small and medium-sized facilities (e.g., those operating on only a single shift which may not have sufficient staff time or expertise to maintain instrumentation). Data reliability remains a concern at facilities of all sizes, and even in facilities where sensors are already deployed, there is a desire for improved operationalizing of the data.

There is therefore a critical role to be played through continued innovation in sensor hardware and commercialization of new technologies – a role that may be shared by academia and industry or potentially led by forward-looking hardware companies or treatment facilities. For all sensor types, there is a clear need for continued improvements toward faster response times, long-term stability, and accuracy under operating conditions (i.e., lower detection limits, improved selectivity, improved stability).More immediately, there may be opportunities for improved operationalizing of the data through use of more advanced data processing approaches, e.g., machine learning or statistical techniques like those discussed above, while hardware innovation progresses.

Taken together, however, these issues point to several more general and fundamental ways in which stakeholders (collectively: academic researchers, practitioners, operators, consulting engineers, educators, etc.) can overcome roadblocks and accelerate innovation in this field.

5.1 Innovation transfer: ensuring research results are relevant to stakeholders

The reality is that innovative process results are exciting because they remain the exception rather than the rule. There may be a number of reasons for the outstanding gap between real-time sensing techniques studied or demonstrated in research labs and those in use in operating treatment facilities. In some facilities, there may be a preference to keep control systems relatively simple or a lack of operational staff trained for this role. In many cases sensors may not yet be sufficiently reliable for deployment in wastewater systems – or there may be a perception that this is the case due to experiences with prior products. Costs may be (or may be perceived as) prohibitive, particularly given the uncertainties surrounding sensor functionality in situ.

A less acknowledged issue is the lack of information in scientific publications that allows operators to make an informed cost–benefit analysis on new technologies. Designing experiments and studies such that one is able to report results as improvement relative to the current approach is critical – this allows operators to directly calculate cost savings or understand potential effluent quality improvements. A related issue is that of lack of consistent or clear reporting metrics across the scientific literature. Among the publications cited in this manuscript, many metrics (MSE, RMSE, percent error, R2, etc.) were reported without units or characterization of the underlying data (e.g., range), making it difficult (or impossible) to evaluate utility in a real application. Studies with wide concentration ranges would be more informative if accuracy and precision metrics were reported both for the overall fit as well as for the range of relevance to plant operations, particularly as the trend to report results only for a multiple-orders-of-magnitude range can artificially inflate the perceived utility if only the lowest end of the concentration range is of relevance to real wastewaters.

Finally, in many studies leveraging logged plant data, there is a tendency to utilize variables that have highest likelihood to be correlated to the target – regardless of whether these are actually measurable online. While the results of such studies may not be usable by operating plants for near-term solutions, they do present an excellent roadmap for directing sensor development goals to support future automation (see the following section).

5.2 Partnerships, data sharing, and accelerating innovation

Development of strong and collaborative partnerships between operators and academic researchers has the potential to be invaluable in testing new sensors and data methods under realistic operating conditions, e.g., as demonstrated already by research partnerships being cultivated at facilities like Hampton Roads,165 the University of Arizona Water & Energy Sustainable Technology (WEST) Center developed in collaboration with Pima County Wastewater Reclamation,166 the NSF ReNUWIt Engineering Research Center Testbeds that allow researchers to demonstrate technologies in pilot-scale plants,167 and the University–Utility Collaborative Partnerships initiated by the Water Environment Federation (WEF).168 The need for stronger partnerships is actually a theme, in our experience, repeated by operators consistently in focus groups and conference sessions, regardless of the specific topic being discussed.

Unfortunately currently such initiatives are rare, regional, and frequently missing representatives from small treatment facilities whose challenges may be very different from large and well-funded districts. Ad hoc collaborative projects (e.g., collection of samples to support a student thesis project) may be more frequent but are less likely to lead to deep or long-term collaborations. This review highlights several instances of data sharing that have enabled development of novel analytical approaches and, importantly, evaluation of these approaches with respect to the status quo. In similar ways, even exploratory data sharing agreements between plants and researchers may be a fruitful avenue for identifying patterns in process chemistry or biology, even with a starting goal of taking fuller advantage of equipment that is already installed at operating facilities (i.e., low cost with potentially large benefit).

Another set of partnerships that may bear fruit is between software and hardware teams, as alluded above. Evidence shows that development of products that integrate hardware and software into soft sensors is likely to become the norm, however currently research on sensor hardware and algorithm development for measurement of nutrients remain relatively separated. Integrated efforts, for instance where results of algorithm development studies (like those presented in section 4.2) suggest priority targets for sensor development or where hardware limitations suggest opportunities for algorithm development, have the potential to vastly accelerate commercialization of high-impact monitoring tools for wastewater applications.

Finally, encouraging adoption of soft sensor instruments – especially those based on complex machine learning architectures – is going to require close collaborations between developers and practitioners to ensure the robustness of the algorithms across the highly variable conditions experienced in individual treatment facilities. Ultimately, as for any machine learning application, this will come down to an issue of data representativeness, and researchers can best understand whether (or to what extent) data are representative through strong and communicative connections with the operators who are familiar with the full range of plant conditions (good and bad) over the course of years. Collaboration throughout the process will enable all stakeholders and partners to share responsibility for assessing the trade-offs between increased model complexity (improved monitoring/prediction accuracy) versus increased number of failure modes, potentially in the limit beyond what can be exhaustively tested. This also highlights the practical need to evaluate how gracefully algorithms fail and build in an ability for self-diagnosis (e.g., recognizing a malfunctioning sensor or influent conditions out of the range for which algorithm is well tuned). Currently the academic literature fails for the most part to address these issues, however collaborative teams have the tools and knowledge required to bridge this gap.

5.3 Expanding innovative solutions to all treatment facilities

In addition to the collaboration building advocated above, it is equally important to highlight the role of partnerships between plant operators, e.g., forums for sharing experiences with specific sensor hardware, ways of interpreting and operationalizing data streams, context and data surrounding infrequent failure modes, and using visualization tools for assessing information from real-time and lab-generated data in an integrated way. In particular, sharing of data, which can only achieve statistically relevant predictive power when aggregated, is a critical component of accelerating adaptation in not just the large and well-funded facilities but also to those facilities serving smaller and potentially less wealthy communities. Some regional conferences, e.g. New England Water Environmental Association (NEWEA) annual and regional meetings, have already begun this process, but round-table discussions and tutorial-style workshops are critically needed to effect meaningful information transfer.

Another potentially rich resource would be a shared data portal, wherein researchers and operators can access historical data from multiple different plants – those using similar processes but at different scales, for example. This would allow for improved identification of patterns (e.g., precursor conditions to common failures) and robustness testing for newly developed soft sensor or control algorithms. While not without challenges (e.g., those of data ownership or privacy concerns, question of who would manage the cyberinfrastructure), such solutions are not without precedent (e.g., for medical diagnostics on rare conditions) and therefore should be closely considered, particularly by professional societies and research organizations supporting this discipline (e.g., the Water Environment Federation, Water Research Foundation).

5.4 Wastewater workforce of tomorrow

Finally, as a growing number of facilities are adopting advanced instrumentation, and retirements will require replacement of nearly 10% of the existing workforce between 2016–2026,169 there is a great need for – as well as an opportunity in – re-envisioning the fundamental training of environmental engineers. Repeatedly, in regional conferences attended by the authors, workshops hosted by the authors, and through global organizations,170,171 operators are stating a growing need for a deeper knowledge of statistics, data processing, data visualization, and data management on their teams. These skills are critical to leverage information from an increasing number of sensors for assessing plant health and automating plant operations. Yet these topics remain minimally addressed in current ABET curriculum requirements172 and therefore most undergraduate programs. Creative adaptation of modules from other disciplines to teach these skills in the context of environmental engineering applications is critical, and design of new curriculum modules or student projects in collaboration with plant operators provides an opportunity to give students insight into the challenges associated with real data. Development of associated practical short-courses, e.g., to be offered through regional professional societies, is also an excellent option for supplementing on-the-job training with a more rigorous and in-depth discussion of these topics.

6 Conclusions

To achieve mandated discharge limits and improve process efficiencies online sensing and automated control are increasingly being implemented for wastewater treatment processes. These needs are amplified by demands related to next-generation processes and environmental sustainability challenges. Real successes have already been shown in improving process management through new sensor-based control schemes such as ABAC (∼10% reduction in energy demand), but highly instrumented WWTPs remain the exception. Academic exploration of novel use of sensor data, e.g., applying machine learning methods, shows the potential for accurate quantification of key analytes (e.g., COD and BOD) at lowered cost and improved accuracy, and development of new control strategies may further improve process results (e.g., nutrient removal and energy efficiencies). Ongoing research in development of novel sensor hardware is pushing the envelope from another direction, for measurement of organic carbon, metals, and even emerging contaminants in wastewaters, though many are not yet ready for commercialization. Major gaps also remain in affordable real-time sensors for phosphate and improved sensors for nitrate/nitrite and ammonia/ammonium (i.e., lower detection limits, better selectivity), as well as generally adoptable sensors for BOD, COD, and TOC (i.e., lower cost, less site-specific calibration). To accelerate overcoming these outstanding challenges, and migration of innovative research from the lab to operating plants, a case is made for building stronger collaborations between operating wastewater treatment facilities and academic research labs, including supporting technology development and technology transfer, ensuring data integrity by design, and working toward an improved environmental engineering curriculum that adequately prepares engineers for this changing landscape in plant design and operations.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We would like to express our gratitude to Mickey Nowak (Massachusetts Water Environment Association) who assisted with planning and logistics for roundtable workshops and to the following individuals who shared with us their experiences to support development of this manuscript: Bob Amaral (Woodard & Curran), Marc Drainville (GHD Group), Brad Furlon (Hoosac Water Quality District), Jeff Gamelli (City of Westfield Water Pollution Control Plant), Tim Hutchins (HACH), Jeff Kalmes (Billerica Water Resource Recovery Facility), Jim Legg (Uxbridge Public Works Department), Tim Loftus (Upper Blackstone Water Pollution Abatement District), Jeff Murawski (City of Fitchburg Water/Wastewater Commission), Jim Nyberg (Sewer Department, Town of North Brookfield), Zack Ritzer (Chicopee Water Department), Edris Taher (Upper Blackstone Water Pollution Abatement District), Liz Tagleiri (Charles River Pollution Control District), Carl Williams (Easthampton Wastewater Treatment Plant), Mike Williams (Holyoke Water Works).

Notes and references

  1. Wastewater treatment, ed. D. H. F. Liu and B. G. Liptak, US: CRC Press, Boca Raton, FL, 2000 Search PubMed.
  2. Clean Water Act (Federal Water Pollution Control Act) [homepage on the Internet]: Washinbgton: US Congress, 1972, Available from: https://www.epa.gov/sites/production/files/2017-08/documents/federal-water-pollution-control-act-508full.pdf Search PubMed.
  3. The EU water framework directive, https://ec.europa.eu/environment/water/water-framework/index_en.html, 2000 Search PubMed.
  4. National Pollutant Discharge Elimination System Regulatory History [homepage on the Internet], Washington, D.C.: EPA, 2018, Available from: https://www.epa.gov/npdes/npdes-regulatory-history Search PubMed.
  5. L. Rieger, S. Gillot, G. Langergraber, T. Ohtsuki, A. Shaw, I. Takacs and S. Winkler, Guidelines for using activated sludge models, IWA publishing, 2012 Search PubMed.
  6. Report on the Performance of Secondary Treatment Technology [homepage on the Internet], Washington, D.C.: EPA, 2013, Available from: https://www3.epa.gov/npdes/pubs/npdes_secondarytreatment_report_march2013.pdf Search PubMed.
  7. A. Joss, N. Derlon, C. Cyprien, S. Burger, I. Szivak and J. Traber, et al., Combined nitritation-anammox: Advances in understanding process stability, Environ. Sci. Technol., 2011, 45(22), 9735–9742 CrossRef CAS.
  8. W. Abma, W. Driessen, R. Haarhuis and M. Van Loosdrecht, Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater, Water Sci. Technol., 2010, 61(7), 1715–1722 CrossRef CAS.
  9. Y. Kong, Y. Xia, J. L. Nielsen and P. H. Nielsen, Ecophysiology of a group of uncultured Gammaproteobacterial glycogenaccumulating organisms in full-scale enhanced biological phosphorus removal wastewater treatment plants, Environ. Microbiol., 2006, 8(3), 479–489 CrossRef CAS.
  10. S. R. Corsi, Evaluation of Energy Conservation Measures for Wastewater Treatment Facilities, Washington, DC: U.S. Environmental Protection Agency, 2010, EPA 832-R-10-005. Steven R. Corsi. Evaluation of energy conservation measures for wastewater treatment facilities, https://www.epa.gov/sites/production/files/2016-01/documents/p1008sbm.pdf, 2010 Search PubMed.
  11. K. Mizuta and M. Shimada, Benchmarking energy consumption in municipal wastewater treatment plants in Japan, Water Sci. Technol., 2010, 62(10), 2256–2262 CrossRef CAS.
  12. K. I. Cogert, R. M. Ziels and M. K. Winkler, Reducing Cost and Environmental Impact of Wastewater Treatment with Denitrifying Methanotrophs, Anammox, and Mainstream Anaerobic Treatment, Environ. Sci. Technol., 2019, 53(21), 12935–12944 CrossRef CAS.
  13. J. T. Bunce, E. Ndam, I. D. Ofiteru, A. Moore and D. W. Graham, A review of phosphorus removal technologies and their applicability to small-scale domestic wastewater treatment systems, Front. Environ. Sci., 2018, 6, 8 CrossRef.
  14. D. Puyol, D. J. Batstone, T. Hülsen, S. Astals, M. Peces and J. O. Krömer, Resource recovery from wastewater by biological technologies: opportunities, challenges, and prospects, Front. Microbiol., 2017, 7, 2106 Search PubMed.
  15. J. Fulcher, Changing the Terms: Water resource recovery facility terminology gains momentum [database on the Internet], Alexandria, VA: Water Environment Federation, 2014, Available from: https://news.wef.org/changing-the-terms/ Search PubMed.
  16. J. C. Kabouris, Modeling, Instrumentation, Automation, and Optimization of Wastewater Treatment Facilities, Water Environ. Res., 1999, 71(5), 729–736,  DOI:10.2175/106143099x133758 , ISSN 10614303.
  17. J. D. Louis Salguero and H. Hohnson, Wastewater flow measurement [database on the Internet], Washington D.C.: U.S. Environmental Protection Agency, 2015, Available from: https://www.epa.gov/sites/production/files/2015-10/documents/wastewater_flow_measurement109_af.r4.pdf Search PubMed.
  18. G. Duffy and F. Regan, Recent developments in sensing methods for eutrophying nutrients with a focus on automation for environmental applications, Analyst, 2017, 142(23), 4355–4372 RSC.
  19. F. Ejeian, P. Etedali, H. A. Mansouri-Tehrani, A. Soozanipour, Z. X. Low and M. Asadnia, et al., Biosensors for wastewater monitoring: A review, Biosens. Bioelectron., 2018, 118, 66–79,  DOI:10.1016/j.bios.2018.07.019 , ISSN 18734235.
  20. E. M. Carstea, J. Bridgeman, A. Baker and D. M. Reynolds, Fluorescence spectroscopy for wastewater monitoring: a review, Water Res., 2016, 95, 205–219 CrossRef CAS.
  21. A. Iratni and N. B. Chang, Advances in control technologies for wastewater treatment processes: status, challenges, and perspectives, IEEE/CAA Journal of Automatica Sinica, 2019, 60(2), 337–363 Search PubMed.
  22. L. Corominas, M. Garrido-Baserba, K. Villez, G. Olsson, U. Cortés and M. Poch, Transforming data into knowledge for improved wastewater treatment operation: A critical review of techniques, Environ. Model. Softw., 2018, 106, 89–103 CrossRef.
  23. P. van Daal, G. Gruber, J. Langeveld, D. Muschalla and F. Clemens, Performance evaluation of real time control in urban wastewater systems in practice: Review and perspective, Environ. Model. Softw., 2017, 95, 90–101 CrossRef.
  24. K. B. Newhart, R. W. Holloway, A. S. Hering and T. Y. Cath, Data-driven performance analyses of wastewater treatment plants: A review, Water Res., 2019, 498–513 CrossRef CAS.
  25. Q. Jiang, X. Yan and B. Huang, Review and perspectives of data-driven distributed monitoring for industrial plant-wide processes, Ind. Eng. Chem. Res., 2019, 58(29), 12899–12912 CrossRef CAS.
  26. F. A. Souza, R. Araújo and J. Mendes, Review of soft sensor methods for regression applications, Chemom. Intell. Lab. Syst., 2016, 152, 69–79 CrossRef CAS.
  27. H. Haimi, M. Mulas, F. Corona and R. Vahala, Data-derived soft-sensors for biological wastewater treatment plants: An overview, Environ. Model. Softw., 2013, 47, 88–107 CrossRef.
  28. K. Solon, E. I. Volcke, M. Spérandio and M. C. van Loosdrecht, Resource recovery and wastewater treatment modelling, Environ. Sci.: Water Res. Technol., 2019, 5(4), 631–642 RSC.
  29. Z. Ge, Review on data-driven modeling and monitoring for plant-wide industrial processes, Chemom. Intell. Lab. Syst., 2017, 171, 16–25 CrossRef CAS.
  30. F. Corona, M. Mulas, H. Haimi, L. Sundell, M. Heinonen and R. Vahala, Monitoring nitrate concentrations in the denitrifying postfiltration unit of a municipal wastewater treatment plant, J. Process Control, 2013, 23(2), 158–170 CrossRef CAS.
  31. G. Olsson, M. Nielsen, Z. Yuan, A. Lynggaard-Jensen and J. P. Steyer, Instrumentation, control and automation in wastewater systems, IWA publishing, 2005 Search PubMed.
  32. P. B. Tchounwou, C. G. Yedjou, A. K. Patlolla and D. J. Sutton, Heavy metal toxicity and the environment, in Molecular, clinical and environmental toxicology, Springer, 2012, pp. 133–164 Search PubMed.
  33. J. Mercer, Wastewater basics for a growing craft brewery [database on the Internet], CBB Media, 2014, Available from: https://www.craftbrewingbusiness.com/equipment-systems/wastewater-basicsgrowing-craft-brewery/ Search PubMed.
  34. J. Umetsu, Instrumentation, control and sensor technology for sewage treatment plants, Fuji Electr. Rev., 1999, 45(4), 126–130 Search PubMed.
  35. M. Tomczak, Final Report: Real-Time, In-Line Sensor for Wastewater Monitoring [database on the Internet], Environmental Protection Agency, 2012, Available from: https://cfpub.epa.gov/ncer_abstracts/index.cfm/fuseaction/display.abstractDetail/abstract/9649/report/F Search PubMed.
  36. W. A. Telliard, Overview of on-line sensors who, what, why [database on the Internet], Environmental Protection Agency, 2015, Available from: https://www.epa.gov/sites/production/files/2015-11/documents/c15-telliard-p1.pdf Search PubMed.
  37. L. Fortuna, S. Graziani, A. Rizzo and M. G. Xibilia, Soft sensors for monitoring and control of industrial processes, Springer Science & Business Media, 2007 Search PubMed.
  38. R. Luttmann, D. G. Bracewell, G. Cornelissen, K. V. Gernaey, J. Glassey and V. C. Hass, et al., Soft sensors in bioprocessing: a status report and recommendations, Biotechnol. J., 2012, 7(8), 1040–1048 CrossRef CAS.
  39. P. Kadlec, B. Gabrys and S. Strandt, Data-driven soft sensors in the process industry, Comput. Chem. Eng., 2009, 33(4), 795–814 CrossRef CAS.
  40. I. Santín, C. Pedret and R. Vilanova, Control and decision strategies in wastewater treatment plants for operation improvement, Springer, 2016, vol. 86 Search PubMed.
  41. B. Petersen, K. Gernaey, J. P. Ottoy and P. Vanrolleghem, Application of biosensors in wastewater treatment, in Proceedings 22e Symposium sur les eaux usées et 11e atelier sur l'eau potable, Montréal, Canada, October 20-21, 1999, 1999, pp. 204–210 Search PubMed.
  42. J. Keskitalo and K. Leiviskä, Application of evolutionary optimisers in data-based calibration of Activated Sludge Models, Expert Syst. Appl., 2012, 39(7), 6609–6617,  DOI:10.1016/j.eswa.2011.12.041.
  43. A. Stare, N. Hvala and D. Vrečko, Modeling, identification, and validation of models for predictive ammonia control in a wastewater treatment plant - A case study, ISA Trans., 2006, 45(2), 159–174 CrossRef CAS.
  44. P. Ingildsen, Realising full-scale control in wastewater treatment systems using in situ nutrient sensors, PhD thesis, Department of Industrial Electrical Engineering and Automation, 2002 Search PubMed.
  45. T. Connor, Case Studies on Implementing Low-Cost Modifications to Improve Nutrient Reduction at Wastewater Treatment Plants [database on the Internet], Environmental Protection Agency, 2015, Available from: https://www.epa.gov/nutrient-policydata/case-studies-implementing-low-costmodifications-improve-nutrient-reduction Search PubMed.
  46. B. Wett, S. Podmirseg, M. Gómez-Brandón, M. Hell, G. Nyhuis and C. Bott, et al., Expanding DEMON sidestream deammonification technology towards mainstream application, Water Environ. Res., 2015, 87(12), 2084–2089 CrossRef CAS.
  47. A. Al-Omari, B. Wett, I. Nopens, H. De Clippeleir, M. Han and P. Regmi, et al., Model-based evaluation of mechanisms and benefits of mainstream shortcut nitrogen removal processes, Water Sci. Technol., 2015, 71(6), 840–847 CrossRef CAS.
  48. T. Lotti, R. Kleerebezem, Z. Hu, B. Kartal, M. De Kreuk and C. van Erp Taalman Kip, et al., Pilot-scale evaluation of anammox-based mainstream nitrogen removal from municipal wastewater, Environ. Technol., 2015, 36(9), 1167–1177 CrossRef CAS.
  49. N. Uri, S. Eriksen, P. H. Nielsen, M. H. Andersen, S. Hafner and Z. Li, et al., Continuous aeration control to reduce N2O emissions in a full-scale sidestream deammonification reactor, Water Environment Federation Technical Exhibition and Conference 2017, WEFTEC 2017, 2017, vol. 2, pp. 1211–1219 Search PubMed.
  50. GENOCOV, Pioneering Technology Could Transform Treatment Plants into Biorefineries [database on the Internet], GENOCOV, 2017, Available from: http://www.genocov.com/who-we-are/news/succesful-inaguration-of-the-sceppharpilot-plant-in-manresa-wwtp Search PubMed.
  51. HACH, Intellical™ ISENH3181 Ammonia (NH3) Ion Selective Electrode (ISE), 1 m cable [database on the Internet], HACH, 2019, Available from: https://www.hach.com/intellical-isenh3181-ammonia-nhsub-3-sub-ion-selective-electrode-ise-1-mcable/product?id=7640512245 Search PubMed.
  52. HACH, Nitratax sc Nitrate Sensors [database on the Internet], HACH, 2019, Available from: https://www.hach.com/nitrate-sensors/nitratax-scnitrate-sensors/family?productCategoryId=35546907021 Search PubMed.
  53. YSI, Pro Series 1004 Ammonium ISE, NH4+ Sensor [database on the Internet], YSI, 2011, Available from: https://www.ysi.com/Product/id-605104/Pro-Series-1004-Ammonium-ISE-NH4-Sensor Search PubMed.
  54. F. Yang, J. Kozak and H. Zhang, Shortcut Biological Nitrogen Removal Methodologies: Mainstream Partial Nitritation/Deammonification and Nitritation/Denitritation, Chicago, IL: Metropolitan Water Reclamation District of Greater Chicago, 2014, Report No.: 14-30 Search PubMed.
  55. O. Larriba, E. Rovira-Cal, Z. Juznic-Zonta, A. Guisasola and J. A. Baeza, Evaluation of the integration of P recovery, polyhydroxyalkanoate production and short cut nitrogen removal in a mainstream wastewater treatment process, Water Res., 2020, 172, 115474 CrossRef CAS.
  56. M. Talebizadeh, E. Belia and P. A. Vanrolleghem, Influent generator for probabilistic modeling of nutrient removal wastewater treatment plants, Environ. Model. Softw., 2016, 77, 32–49 CrossRef.
  57. M. O'Shaughnessy, Mainstream Deammonification, Water Environment Research Foundation Alexandria, VA, 2015 Search PubMed.
  58. K. Uprety, Implementing and operating ammonia-based aeration control at a biological nutrient removal facility [dissertation], PhD thesis, Old Dominion University, Virginia Beach, VA, 2014 Search PubMed.
  59. B. Dabkowski, R. Relph, G. Wendorf, K. Menning and M. White, Modelling Ammonia Based Aeration Control in Real Time with Online Instrumentation, Proceedings of the Water Environment Federation, 2017, 2017(10), 2996–3001 CrossRef.
  60. HACH, Case study: pilots Ammonia-Based Aeration Control with Hach RTC-N to optimize nitrification [database on the Internet], HACH, 2016, Available from: https://www.hach.com/asset-get.download-en.jsa%3Fcode%3D230518+&cd=1&hl=en&ct=clnk&gl=us Search PubMed.
  61. O. Schraa, L. Rieger, J. Alex and I. Miletić, Ammonia-based aeration control with optimal SRT control: improved performance and lower energy consumption, Water Sci. Technol., 2019, 79(1), 63–72 CrossRef.
  62. B. F. Smets, C. P. i Nàcher, C. D. Felez, M. M. Jensen, E. Ramin and B. G. Plósz, et al., Modelling N2O dynamics in the engineered N cycle: Evaluation of alternate model structures, in 4th IWA/WEF Wastewater Treatment Modelling Seminar, 2014, Water Environment Federation, 2014, pp. 343–346 Search PubMed.
  63. E. Taher, K. Sangrey and T. Loftus, A Multi-Sector Approach to Optimize Process Efficiency and Reduce Operating Costs at the Upper Blackstone [Conference Presentation], NEWEA Annual conference, Boston, MA, 2020 Search PubMed.
  64. J. Gamelli and K. Gagnon, Making the Most of Online Instrumentation to Improve Process Performance and Reduce Costs at the Westfield, MA [Conference Presentation], NEWEA Annual conference, Boston, MA, 2020 Search PubMed.
  65. A. Li, How Big Data enables smart collection systems and protect Wastewater Treatment Plants [database on the Internet], Wastewater Management Solutions, 2019, Available from: https://www.kando.eco/how-canbig-data-help-protect-wastewater-treatmentplants-environment Search PubMed.
  66. Water-Tech W. 9 Breakthrough Technologies Transforming the Water-Tech Industry [database on the Internet], World Water-Tech, 2020, Available from: https://worldwatertechnorthamerica.com/7-breakthrough-technologies-transforming-thewater-tech-industry/ Search PubMed.
  67. B. W. Bequette, Process control: modeling, design, and simulation, Prentice Hall Professional, 2003 Search PubMed.
  68. T. E. Marlin, Process Control, Designing Processes and Control Systems for Dynamic Performance, McGraw-Hill Science/Engineering/Math, 2002 Search PubMed.
  69. R. Tzoneva, Optimal PID control of the dissolved oxygen concentration in the wastewater treatment plant, in AFRICON, 2007, IEEE, 2007, pp. 1–7 Search PubMed.
  70. N. A. Wahab, R. Katebi and J. Balderud, Multivariable PID control design for activated sludge process with nitrification and denitrification, Biochem. Eng. J., 2009, 45(3), 239–248 CrossRef CAS.
  71. C. Yoo, J. Cho, H. Kwak, S. Choi, H. Chun and I. Lee, Closedloop identification and control application for dissolved oxygen concentration in a full-scale coke wastewater treatment plant, Water Sci. Technol., 2001, 43(7), 207–214 CrossRef CAS.
  72. M. Drahansky, M. T. Paridah, A. Moradbak, A. Z. Mohamed, F. Abdulwahab taiwo Owolabi and M. Asniza, et al., Distributed Control Systems for a Wastewater Treatment Plant: Architectures and Advanced Control Solutions, Intech, 2016, i(tourism):13. Available from: https://www.intechopen.com/books/advancedbiometric-technologies/liveness-detectionin-biometrics Search PubMed.
  73. B. H. Kiepper, Understanding Laboratory Wastewater Tests Series, The University of Georgia, 2016, Available from: https://secure.caes.uga.edu/extension/publications/files/pdf/C%20992_3.PDF.
  74. EPA, Analytical Methods for the Determination of Pollutants in Pharmaceutical Manufacturing Industry Wastewater, EPA, Office of Water, Engineering and Analysis Division, 1995, Available from: https://books.google.com/books?id=-kdSiK27pAAC Search PubMed.
  75. B. Drosg, R. Braun, G. Bochmann and T. Al Saedi, Analysis and characterisation of biogas feedstocks, in The Biogas Handbook, Elsevier, 2013, pp. 52–84 Search PubMed.
  76. P. Whitehead, Total Organic Carbon (TOC) and its measurement [database on the Internet], VWS, 2018, Available from: https://www.elgalabwater.com/blog/total-organic-carbon-toc Search PubMed.
  77. T. Wright, inventor; ClearCove Systems Inc, assignee, Automatic control of biochemical oxygen demand content (bod) of sludge and other products of a waste water treatment system, 2019 Sep 21 Search PubMed.
  78. A. Guerrini, G. Romano and A. Indipendenza, Energy efficiency drivers in wastewater treatment plants: a double bootstrap DEA analysis, Sustainability, 2017, 9(7), 1126 CrossRef.
  79. B. Dweik, K. Harrison and A. Argun, Rapid Determination of Total Organic Carbon (TOC) in Water Systems, in 49th International Conference on Environmental Systems, 2019, p. 10 Search PubMed.
  80. N. Al-Dasoqi, A. Mason, R. Alkhaddar and A. Al-Shamma'a, Use of sensors in wastewater quality monitoring—a review of available technologies, in World Environmental and Water Resources Congress 2011, Bearing Knowledge for Sustainability, 2011, pp. 3379–3388 Search PubMed.
  81. O. Korostynska, A. Mason and A. Al-Shamma'a, Monitoring of nitrates and phosphates in wastewater: current technologies and further challenges, Int. J. Smart Sens. Intell. Syst., 2012, 5(1) DOI:10.21307/ijssis-2017-475.
  82. N. H. Tran, M. Reinhard and K. Y. H. Gin, Occurrence and fate of emerging contaminants in municipal wastewater treatment plants from different geographical regions-a review, Water Res., 2018, 133, 182–207 CrossRef CAS.
  83. X. Chen, A. T. Mielczarek, K. Habicht, M. H. Andersen, D. Thornberg and G. Sin, Assessment of Full-Scale N2O Emission Characteristics and Testing of Control Concepts in an Activated Sludge Wastewater Treatment Plant with Alternating Aerobic and Anoxic Phases, Environ. Sci. Technol., 2019, 53(21), 12485–12494 CrossRef CAS.
  84. H. Kosonen, M. Heinonen, A. Mikola, H. Haimi, M. Mulas and F. Corona, et al., Nitrous oxide production at a fully covered wastewater treatment plant: results of a long-term online monitoring campaign, Environ. Sci. Technol., 2016, 50(11), 5547–5554 CrossRef CAS.
  85. H. Lu and K. Chandran, Factors promoting emissions of nitrous oxide and nitric oxide from denitrifying sequencing batch reactors operated with methanol and ethanol as electron donors, Biotechnol. Bioeng., 2010, 106(3), 390–398 CAS.
  86. P. A. Vanrolleghem and D. S. Lee, On-line monitoring equipment for wastewater treatment processes: state of the art, Water Sci. Technol., 2003, 47(2), 1–34 CrossRef CAS.
  87. W. Bourgeois, J. E. Burgess and R. M. Stuetz, On-line monitoring of wastewater quality: a review, J. Chem. Technol. Biotechnol., 2001, 76(4), 337–348 CrossRef CAS.
  88. C. Warwick, A. Guerreiro and A. Soares, Sensing and analysis of soluble phosphates in environmental samples: A review, Biosens. Bioelectron., 2013, 41, 1–11 CrossRef CAS.
  89. G. Zhou, B. Jin, Y. Wang, Q. Dong, A. Maity and J. Chang, et al., Ultrasensitive sensors based on aluminum oxide-protected reduced graphene oxide for phosphate ion detection in real water, Mol. Syst. Des. Eng., 2020, 936–942 RSC.
  90. S. R. Corsi, Using optical sensors to detect sewage contamination in the Great Lakes [database on the Internet], USGS, 2016, Available from: https://www.usgs.gov/centers/umid-water/science/using-opticalsensors-detect-sewage-contamination-greatlakes?qt-science_center_objects=0#qtscience_center_objects Search PubMed.
  91. G. M. Ferrari, M. D. Dowell, S. Grossi and C. Targa, Relationship between the optical properties of chromophoric dissolved organic matter and total concentration of dissolved organic carbon in the southern Baltic Sea region, Mar. Chem., 1996, 55(3–4), 299–316 CrossRef.
  92. D. M. Reynolds, The differentiation of biodegradable and nonbiodegradable dissolved organic matter in wastewaters using fluorescence spectroscopy, J. Chem. Technol. Biotechnol., 2002, 77(8), 965–972 CrossRef CAS.
  93. A. Goffin, S. Guérin, V. Rocher and G. Varrault, Towards a better control of the wastewater treatment process: excitationemission matrix fluorescence spectroscopy of dissolved organic matter as a predictive tool of soluble BOD 5 in influents of six Parisian wastewater treatment plants, Environ. Sci. Pollut. Res., 2018, 25(9), 8765–8776 CrossRef CAS.
  94. A. Goffin, S. Guérin-Rechdaoui, V. Rocher and G. Varrault, An environmentally friendly surrogate method for measuring the soluble chemical oxygen demand in wastewater: use of three-dimensional excitation and emission matrix fluorescence spectroscopy in wastewater treatment monitoring, Environ. Monit. Assess., 2019, 191(7), 421 CrossRef.
  95. s::can. spectro::lyser [database on the Internet]. s::can. 2019. Available from: https://www.s-can.at/products/spectrometer-probes.
  96. G. Systemelektronik, Application of the ISA Spectrometer in a wastewater treatment plant [database on the Internet], GO Systemelektronik GmbH, 2019, Available from: https://www.go-sys.de/en/applications/wastewater/ Search PubMed.
  97. Eijkelkamp, SPECTRAL-8 (UV/VIS) [database on the Internet], Eijkelkamp, 2017, Available from: https://www.eijkelkamp-usa.com/news/eijkelkampspectrometer-uv-vis-news.htm Search PubMed.
  98. Realtech, A dairy plant in Minnesota was reliant on daily BOD and COD grab samples for detection of product loss [database on the Internet], Realtech, 2018, Available from: https://realtechwater.com/w_p/wpcontent/uploads/2019/10/Minnesota-Dairy-Product-Loss-Real-Tech-Inc.pdf Search PubMed.
  99. M. Saif, R. Kamal and H. Hafez, BaZrO3: Eu3+ nanophosphor: A potential fluorescent sensor for highly selective and sensitive detection of chromium ions from tannery leather and wastewater, J. Alloys Compd., 2019, 803, 658–663 CrossRef CAS.
  100. M. Shenashen, S. El-Safty and E. Elshehy, Architecture of optical sensor for recognition of multiple toxic metal ions from water, J. Hazard. Mater., 2013, 260, 833–843 CrossRef CAS.
  101. N. Sivchenko, K. Kvaal and H. Ratnaweera, Floc sensor prototype tested in the municipal wastewater treatment plant, Cogent Eng., 2018, 5(1), 1436929 Search PubMed.
  102. S. Amemiya, Potentiometric ion-selective electrodes, in Handbook of electrochemistry, Elsevier, 2007, pp. 261–294 Search PubMed.
  103. E. Bakker, Ion-Selective Electrodes: Overview, Encyclopedia of Analytical Science, 2019, pp. 231–251 Search PubMed.
  104. K. N. Mikhelson, Ion-selective electrodes, Springer, 2013, vol. 81 Search PubMed.
  105. A. M. Snauffer, U. Chauhan, K. Cogert, M. K. Winkler and A. V. Mueller, Data Fusion for Environmental Process Control: Maximizing Useful Information Recovery under Data Limited Constraints, IEEE Sens. Lett., 2018, 3(1), 1–4 Search PubMed.
  106. K. V. Gupta, A. Nayak, S. Agarwal and B. Singhal, Recent advances on potentiometric membrane sensors for pharmaceutical analysis, Comb. Chem. High Throughput Screening, 2011, 14(4), 284–302 CrossRef.
  107. P. Liu, H. Zhang, Y. Feng, C. Shen and F. Yang, Integrating electrochemical oxidation into forward osmosis process for removal of trace antibiotics in wastewater, J. Hazard. Mater., 2015, 296, 248–255 CrossRef CAS.
  108. H. T. Elbalkiny, A. M. Yehia, M. R. Safa'a and Y. S. Elsaharty, Potentiometric diclofenac detection in wastewater using functionalized nanoparticles, Microchem. J., 2019, 145, 90–95 CrossRef CAS.
  109. D. I. Kim, P. Dorji, G. Gwak, S. Phuntsho, S. Hong and H. Shon, Reuse of municipal wastewater via membrane capacitive deionization using ion-selective polymer-coated carbon electrodes in pilot-scale, Chem. Eng. J., 2019, 372, 241–250 CrossRef CAS.
  110. S. Papias, M. Masson, S. Pelletant, S. P. Boucle and C. Boutin, In situ continuous monitoring of nitrogen with ion-selective electrodes in a free-water surface constructed wetland accepting treated wastewater: an operating protocol to obtain reliable data, in 15th IWA International Conference on Wetland Systems for Water Pollution Control, Systems for Water Pollution Control, 2016, p. 10 Search PubMed.
  111. Y. Huang, T. Wang, Z. Xu, E. Hughes, F. Qian and M. Lee, et al., Realtime in situ monitoring of nitrogen dynamics in wastewater treatment processes using wireless, solid-state, and ionselective membrane sensors, Environ. Sci. Technol., 2019, 53(6), 3140–3148 CrossRef CAS.
  112. J. Ali, J. Najeeb, M. A. Ali, M. F. Aslam and A. Raza, Biosensors: their fundamentals, designs, types and most recent impactful applications: a review, J. Biosens. Bioelectron., 2017, 8(1), 1–9 Search PubMed.
  113. F. Ejeian, P. Etedali, H. A. Mansouri-Tehrani, A. Soozanipour, Z. X. Low and M. Asadnia, et al., Biosensors for wastewater monitoring: A review, Biosens. Bioelectron., 2018, 118, 66–79 CrossRef CAS.
  114. A. Rodriguez-Sanchez, B. Muñoz-Palazon, M. Hurtado-Martinez, A. Mikola, J. Gonzalez-Lopez and R. Vahala, et al., Analysis of microbial communities involved in organic matter and nitrogen removal in a full-scale moving bed biofilm reactor located near the Polar Arctic Circle, Int. Biodeterior. Biodegrad., 2020, 146, 104830 CrossRef CAS.
  115. M. Collivignarelli, R. Pedrazzani, S. Sorlini, A. Abbà and G. Bertanza, H2O2 based oxidation processes for the treatment of real high strength aqueous wastes, Sustainability, 2017, 9(2), 244 CrossRef.
  116. F. Ejeian, P. Etedali, H. A. Mansouri-Tehrani, A. Soozanipour, Z. X. Low and M. Asadnia, et al., Biosensors for wastewater monitoring: A review, Biosens. Bioelectron., 2018, 118, 66–79 CrossRef CAS.
  117. S. Sumaraj and M. Ghangrekar, Development of microbial fuel cell as biosensor for detection of organic matter of wastewater, Recent Res. Sci. Technol., 2014, 162–166 Search PubMed.
  118. A. Adekunle, V. Raghavan and B. Tartakovsky, A comparison of microbial fuel cell and microbial electrolysis cell biosensors for real-time environmental monitoring, Bioelectrochemistry, 2019, 126, 105–112 CrossRef CAS.
  119. M. Nielsen, L. H. Larsen, M. S. Jetten and N. P. Revsbech, Bacteriumbased NO2- biosensor for environmental applications, Appl. Environ. Microbiol., 2004, 70(11), 6551–6558 CrossRef CAS.
  120. S. Zhao, P. Liu, Y. Niu, Z. Chen, A. Khan and P. Zhang, et al., A novel early warning system based on a sediment microbial fuel cell for in situ and real time hexavalent chromium detection in industrial wastewater, Sensors, 2018, 18(2), 642 CrossRef.
  121. C. Schirmer, J. Posseckardt, M. Schröder, M. Gläser, S. Howitz and W. Scharff, et al., Portable and low-cost biosensor towards on-site detection of diclofenac in wastewater, Talanta, 2019, 203, 242–247 CrossRef CAS.
  122. Y. Zou, J. Pisciotta, R. B. Billmyre and I. V. Baskakov, Photosynthetic microbial fuel cells with positive light response, Biotechnol. Bioeng., 2009, 104(5), 939–946 CrossRef CAS.
  123. T. Yamashita, N. Ookawa, M. Ishida, H. Kanamori, H. Sasaki and Y. Katayose, et al., A novel open-type biosensor for the insitu monitoring of biochemical oxygen demand in an aerobic environment, Sci. Rep., 2016, 6(1), 1–9 CrossRef.
  124. G. Pankratova and L. Gorton, Electrochemical communication between living cells and conductive surfaces, Curr. Opin. Electrochem., 2017, 5(1), 193–202 CrossRef CAS.
  125. A. Liu, Q. Lang, B. Liang and J. Shi, Sensitive detection of maltose and glucose based on dual enzyme-displayed bacteria electrochemical biosensor, Biosens. Bioelectron., 2017, 87, 25–30 CrossRef CAS.
  126. A. Saberi, F. Jabbari, P. Zarrintaj, M. R. Saeb and M. Mozafari, Electrically Conductive Materials: Opportunities and Challenges in Tissue Engineering, Biomolecules, 2019, 9(9), 448 CrossRef CAS.
  127. M. M. Hamed, M. G. Khalafallah and E. A. Hassanien, Prediction of wastewater treatment plant performance using artificial neural networks, Environ. Model. Softw., 2004, 19(10), 919–928 CrossRef.
  128. D. Hanbay, I. Turkoglu and Y. Demir, Prediction of wastewater treatment plant performance based on wavelet packet decomposition and neural networks, Expert Syst. Appl., 2008, 34(2), 1038–1043 CrossRef.
  129. Office EP, Multi Sensor Fusion for Real-time Monitoring of Waste Water Quality [database on the Internet], European Commission, 2014, Available from: https://cordis.europa.eu/project/rcn/99331/factsheet/en Search PubMed.
  130. A. V. Mueller and H. F. Hemond, Extended artificial neural networks: incorporation of a priori chemical knowledge enables use of ion selective electrodes for in-situ measurement of ions at environmentally relevant levels, Talanta, 2013, 117, 112–118 CrossRef CAS.
  131. J. Koka, Machine learning makes water smarter [database on the Internet], Water & Wastes Digest, 2019, Available from: https://www.wwdmag.com/machine-learning-0 Search PubMed.
  132. J. Filipe, R. J. Bessa, M. Reis, R. Alves and P. Póvoa, Data-driven predictive energy optimization in a wastewater pumping station, Appl. Energy, 2019, 252, 113423 CrossRef.
  133. A. Amaral and E. Ferreira, Activated sludge monitoring of a wastewater treatment plant using image analysis and partial least squares regression, Anal. Chim. Acta, 2005, 544(1–2), 246–253 CrossRef CAS.
  134. C. Quintelas, D. P. Mesquita, E. C. Ferreira and A. L. Amaral, Quantification of pharmaceutical compounds in wastewater samples by near infrared spectroscopy (NIR), Talanta, 2019, 194, 507–513 CrossRef CAS.
  135. X. Qin, F. Gao and G. Chen, Wastewater quality monitoring system using sensor fusion and machine learning techniques, Water Res., 2012, 46(4), 1133–1144 CrossRef CAS.
  136. M. W. Lee, S. H. Hong, H. Choi, J. H. Kim, D. S. Lee and J. M. Park, Realtime remote monitoring of small-scaled biological wastewater treatment plants by a multivariate statistical process control and neural network-based software sensors, Process Biochem., 2008, 43(10), 1107–1113 CrossRef CAS.
  137. Z. J. Liu, J. Q. Wan, Y. W. Ma and Y. Wang, Online prediction of effluent COD in the anaerobic wastewater treatment system based on PCA-LSSVM algorithm, Environ. Sci. Pollut. Res., 2019, 26(13), 12828–12841 CrossRef CAS.
  138. C. K. Yoo, P. A. Vanrolleghem and I. B. Lee, Nonlinear modeling and adaptive monitoring with fuzzy and multivariate statistical methods in biological wastewater treatment plants, J. Biotechnol., 2003, 105(1–2), 135–163 CrossRef CAS.
  139. M. Zounemat-Kermani, D. Stephan and R. Hinkelmann, Multivariate NARX neural network in prediction gaseous emissions within the influent chamber of wastewater treatment plants, Atmos. Pollut. Res., 2019, 10(6), 1812–1822 CrossRef CAS.
  140. T. da Silva Ribeiro, C. D. Grossi, A. G. Merma, B. F. dos Santos and M. L. Torem, Removal of boron from mining wastewaters by electrocoagulation method: Modelling experimental data using artificial neural networks, Miner. Eng., 2019, 131, 8–13 CrossRef CAS.
  141. P. Antwi, J. Li, J. Meng, K. Deng, F. K. Quashie and J. Li, et al., Feedforward neural network model estimating pollutant removal process within mesophilic upflow anaerobic sludge blanket bioreactor treating industrial starch processing wastewater, Bioresour. Technol., 2018, 257, 102–112 CrossRef CAS.
  142. M. S. Zaghloul, R. A. Hamza, O. T. Iorhemen and J. H. Tay, Performance prediction of an aerobic granular SBR using modular multilayer artificial neural networks, Sci. Total Environ., 2018, 645, 449–459 CrossRef CAS.
  143. H. Gong, Artificial Neural Network Modeling for Organic and Total Nitrogen Removal of Aerobic Granulation [thesis], MPhil thesis, University of Calgary, Alberta, Canada, 2018 Search PubMed.
  144. T. Mehmood, K. H. Liland, L. Snipen and S. Sæbø, A review of variable selection methods in partial least squares regression, Chemom. Intell. Lab. Syst., 2012, 118, 62–69 CrossRef CAS.
  145. D. S. Lee, M. W. Lee, S. H. Woo, Y. J. Kim and J. M. Park, Nonlinear dynamic partial least squares modeling of a full-scale biological wastewater treatment plant, Process Biochem., 2006, 41(9), 2050–2057 CrossRef CAS.
  146. S. H. Woo, C. O. Jeon, Y. S. Yun, H. Choi, C. S. Lee and D. S. Lee, On-line estimation of key process variables based on kernel partial least squares in an industrial cokes wastewater treatment plant, J. Hazard. Mater., 2009, 161(1), 538–544 CrossRef CAS.
  147. H. Xiao, B. Bai, X. Li, J. Liu, Y. Liu and D. Huang, Interval multiple-output soft sensors development with capacity control for wastewater treatment applications: A comparative study, Chemom. Intell. Lab. Syst., 2019, 184, 82–93 CrossRef CAS.
  148. U. Showkat and I. A. Najar, Study on the efficiency of sequential batch reactor (SBR)-based sewage treatment plant, Appl. Water Sci., 2019, 9(1), 2 CrossRef.
  149. Y. Dong and S. J. Qin, A novel dynamic PCA algorithm for dynamic data modeling and process monitoring, J. Process Control, 2018, 67, 1–11 CrossRef CAS.
  150. D. Aguado, T. Montoya, L. Borras, A. Seco and J. Ferrer, Using SOM and PCA for analysing and interpreting data from a P-removal SBR, Eng. Appl. Artif. Intell., 2008, 21(6), 919–930 CrossRef.
  151. H. A. Chipman and H. Gu, Interpretable dimension reduction, J. Appl. Stat., 2005, 32(9), 969–987 CrossRef.
  152. T. Miettinen, T. J. Hurse, M. A. Connor, S. P. Reinikainen and P. Minkkinen, Multivariate monitoring of a biological wastewater treatment process: a case study at Melbourne Water's Western Treatment Plant, Chemom. Intell. Lab. Syst., 2004, 73(1), 131–138 CrossRef CAS.
  153. M. Fan, J. Hu, R. Cao, W. Ruan and X. Wei, A review on experimental design for pollutants removal in water treatment with the aid of artificial intelligence, Chemosphere, 2018, 200, 330–343 CrossRef CAS.
  154. F. Despagne and D. L. Massart, Neural networks in multivariate calibration, Analyst, 1998, 123(11), 157R–178R RSC.
  155. J. F. de Canete, P. del Saz-Orozco, R. Baratti, M. Mulas, A. Ruano and A. Garcia-Cerezo, Soft-sensing estimation of plant effluent concentrations in a biological wastewater treatment plant using an optimal neural network, Expert Syst. Appl., 2016, 63, 8–19 CrossRef.
  156. H. Gong, R. Pishgar and J. Tay, Artificial neural network modelling for organic and total nitrogen removal of aerobic granulation under steady-state condition, Environ. Technol., 2019, 40(24), 3124–3139 CrossRef CAS.
  157. P. Zhou, Z. Li, S. Snowling, R. Goel and Q. Zhang, Short-Term Wastewater Influent Prediction Based on Random Forests and Multi-Layer Perceptron, Journal of Environmental Informatics Letters, 2019, 1(2), 87–93 Search PubMed.
  158. V. Belikova, V. Panchuk, E. Legin, A. Melenteva, D. Kirsanov and A. Legin, Continuous monitoring of water quality at aeration plant with potentiometric sensor array, Sens. Actuators, B, 2019, 282, 854–860 CrossRef CAS.
  159. S. Mazhar, A. Ditta, L. Bulgariu, I. Ahmad, M. Ahmed and A. A. Nadiri, Sequential treatment of paper and pulp industrial wastewater: Prediction of water quality parameters by Mamdani Fuzzy Logic model and phytotoxicity assessment, Chemosphere, 2019, 227, 256–268 CrossRef CAS.
  160. M. Fiter, D. Güell, J. Comas, J. Colprim, M. Poch and I. Rodríguez-Roda, Energy saving in a wastewater treatment process: An application of Fuzzy Logic Control, Environ. Technol., 2005, 26(11), 1263–1270 CrossRef CAS.
  161. G. Langergraber, N. Fleischmann and F. Hofstaedter, A multivariate calibration procedure for UV/VIS spectrometric quantification of organic matter and nitrate in wastewater, Water Sci. Technol., 2003, 47(2), 63–71 CrossRef CAS.
  162. A. Bouzas, N. Martí, S. Grau, R. Barat, D. Mangin and L. Pastor, Implementation of a global P-recovery system in urban wastewater treatment plants, J. Cleaner Prod., 2019, 227, 130–140,  DOI:10.1016/j.jclepro.2019.04.126.
  163. C. Simion, O. Chenaru, G. Florea, J. I. Lozano, S. Nabulsi and M. Reis, et al., Decision support system based on fuzzy control for a wastewater treatment plant, Int. J. Environ. Sci., 2016, 1, 344–349 Search PubMed.
  164. L. Rieger, R. M. Jones, P. L. Dold and C. B. Bott, Ammonia-Based Feedforward and Feedback Aeration Control in Activated Sludge Processes, Water Environ. Res., 2014, 86(1), 63–73 CrossRef CAS.
  165. HRSD, Hampton Roads Treatment Plants 2020 Vision [database on the Internet], HRSD, 2019, Available from: https://www.hrsd.com/sites/default/iles/assets/Documents/StrategicPlan-2020Vision_0.pdf Search PubMed.
  166. WEST, University of Arizona's Water & Energy Sustainable Technology Center [database on the Internet], University of Arizona, 2019, Available from: https://west.arizona.edu/ Search PubMed.
  167. ReNUWIt, Re-inventing the Nation's Urban Water Infrastructure (ReNUWIt) [database on the Internet], ReNUWIt, 2020, Available from: http://renuwit.org/research/ Search PubMed.
  168. Federation WE, University-Utility Collaborative Partnerships [database on the Internet], Water Environment Federation, 2017, Available from: https://www.wef.org/globalassets/assetswef/direct-download-library/public/03---resources/wsec-2017-tr-005-universityutility-collaborative-partnerships.pdf Search PubMed.
  169. Office GA, Water and Wastewater Workforce: Recruiting Approaches Helped Industry Hire Operators, but Additional EPA Guidance Could Help Identify Future Needs [database on the Internet], United States Government Accountability Office, 2018, Available from: https://www.gao.gov/assets/690/689621.pdf Search PubMed.
  170. of Labor Statistics B. Occupational Outlook Handbook [database on the Internet], U.S. Department of Labor, 2020, Available from: https://www.bls.gov/OOH/production/water-and-wastewater-treatmentplant-and-system-operators.htm#tab-6 Search PubMed.
  171. N. Y. M. Rüdiger Heidebrecht, Skills Challenges in the Water and Wastewater Industry [database on the Internet], German Association for Water, Wastewater and Waste, 2012, Available from: https://unevoc.unesco.org/fileadmin/user_upload/docs/WaterBooklet.pdf Search PubMed.
  172. K. J. Fridley, D. B. Hains, A. N. Morse and L. Nolen The CEBOK3 and ABET Accreditation Criteria: A Gap Analysis, in 2019 ASEE Annual Conference & Exposition. ASEE, 2019, p. 22 Search PubMed.

This journal is © The Royal Society of Chemistry 2020