Total ammonia aeration control (TAAC) theory – An innovative ammonia-based aeration controller

There has been much progress in the improvement of controls for aeration in secondary treatment processes at water resource recovery facilities (WRRFs) over recent years. Historically, many WRRFs have adjusted the air supplied to secondary biological nutrient removal (BNR) systems to maintain a dissolved oxygen (DO) content of the process aerated zones, generally around 1.0–2.5 mg/L (Åmand 2014). Energy use from aeration has been vastly researched because it quantifies commonly as half of the consumption of a WRRF (WEF 2009), and some reports show this to be as high as 75% (Reardon 1995). Many case studies of full-scale implementations have shown that ammonia-based aeration control (ABAC) results in increased treatment effectiveness and a substantial reduction of energy consumption from aeration, often 15–25% (Rieger et al. 2012a).

It has been proven that migration to a supervisory controller that is ammonia based will provide cost savings by reducing the DO setpoints to lower levels, versus overaerating the media to ensure full nitrification of the ammonia content. Another benefit is an increase in the overall nitrogen removal through simultaneous nitrification/denitrification (SND) (Jimenez et al. 2013). In this article, I will detail the development of a novel controller for these systems that use the total influent ammonia rate to provide optimal control stability of the primary output variable, ammonia concentration, achieved from the minimal information necessary for the process. This method enables WRRF entities to utilize a real-time data-driven model and precise aeration setpoints, without being encumbered with the additional effort for installation and maintenance of devices required for traditional control approaches.

A brief overview of the purpose of aeration for the conventional secondary treatment process: the activated sludge process uses oxygen as the catalyst for a two-step biological reaction, followed by a second biological reaction that produces nitrogen gas. The two-step reaction includes: (1) nitritation, where ammonia-oxidizing bacteria population converts ammonia to nitrite; and (2) nitrification, where a population of nitrite-oxidizing bacteria converts nitrite to nitrate (Rieger et al. 2013). The nitrification of ammonia to nitrate is the precursor to the denitrification reaction, sometimes required depending on effluent limits, where a population of heterotrophic bacteria will convert the nitrate to nitrogen gas. This gas will dissipate into the atmosphere with the final results of all of the reactions from this process driving a total nitrogen removal rate.

These systems develop an inherit nitrogen removal rate that is primarily determined by the existing nitrifier biomass, set by the average ammonia load and the solids retention time (SRT) (Rieger et al. 2013). This rate is the most direct quantification of efficiency of the treatment process and can be increased significantly via SND. It has been shown that SND is possible via two paths: (1) conventional, reactions mentioned earlier in this section occur concurrently, and (2) nitritation–denitrification (nitrite-shunt), where the nitrification step is inhibited, and denitrification converts nitrite directly to nitrogen gas. It has been shown that this is of particular interest for potential savings, but controlling the DO to the proper setpoint is challenging (Jimenez et al. 2013).

Feedforward controllers have traditionally used established models to predictively generate DO or airflow setpoints (Åmand & Carlsson 2013). The use of the latter will result in less significant airflow changes during operation, which is beneficial by reducing wear on the primary aeration equipment (blowers and valves). In particular, WRRFs that have large spikes of influent ammonia would benefit from utilizing feedforward. For these cases, the plug flow return will have a higher concentration of ammonia and compound these events when using only feedback control strategies. This is because the error deviation from setpoint must occur at the effluent measurement point prior to any corrective action being taken (Rieger et al. 2013). This leads to the question, Could both feedforward and feedback methods be combined into a single controller to derive the benefits of both and negate the pitfalls?

The result of this question is the thesis of this article. Total ammonia aeration control (TAAC) is a four-element controller that uses PID feedback control in conjunction with recent historical data to directly determine airflow setpoints. It leverages statistical analysis to find the current linear relationship between: (1) influent loading, the rate of total ammonia (flow rate × ammonia concentration) entering the treatment area, and (2) the oxygen rate required given the existing system conditions to achieve the desired effluent setpoint. This analytic is often referred to as the specific air of the process, but in general terms, it is the quantification of the process efficiency. The algorithm uses this analytic to initially calculate a feedforward component. The TAAC algorithm then makes an adjustment, via the feedback component, which modifies this calculated real-time air demand to account for disturbances and/or changes that occur over time within the system (e.g., temperature, pH, SRT (Åmand et al. 2013)).

Increased capacity and enhanced performance via optimization of a system through control improvements come at low costs for implementation, especially when compared to alternatives that require physical modifications. As operational excellence efforts are conducted across multiple industries, there have been major strides made in controller design for complex systems, such as aeronautical flight control and wastewater aeration. This section provides an informational brief on how the progression of PID controller design has contributed to major parts of the advancement in the process control domain.

This implies that the use of a DO variable within any PID controller is not the optimal solution to maintain a stable ammonia concentration and will lead to some instability in ABAC implementations. It often results in the need for regular tuning of the PID parameters to adjust for system characteristics to create quality control.

A more advanced implementation is known as a cascade loop, where two or more PID algorithms in series create outer and inner PID functions, which results in a single controller output to a final element. This is the same type of approach commonly used for feedback ABAC solutions. The concept is presented here in the following example of a simpler system, control of the temperature in a large capacitance tank. The inner loop PID provides control of the smaller capacitance heat supply, which will require faster responses, while the outer loop controller provides the inner loop a varied setpoint based on operational conditions. This outer loop PID is the reaction to the error of the large tank temperature from the overall system desired setpoint. The two separate PID functions are able to be tuned with the proper parameters that meet the requirements of the different system properties. A visual depiction of this example is shown in Figure 3.

A three-element controller is useful and sometimes necessary for systems that require precise manipulation. This method utilizes a feedforward aspect to set an input (manipulated) variable expectation and then applies a compensation based on the output from the controlled variable feedback PID controller. Generation of the feedforward component alters the setup of the feedback adjustment loop and enables the use of specifically tuned parameters for the given system, while the final closed-loop PID maintains the stability of the calculated input variable setpoint. This allows for feedforward and feedback factors that can be varied to account for differences in the mechanistic properties of a system. Boiler drum level controllers often use this design.

Previous aeration controllers that included a feedforward aspect have relied on biological modeling to predict the required air using loading inputs and system assumptions. These solutions would still often output a DO setpoint for the control zones and would not be able to account for inherited field device issues or variations over time of the model parameters (e.g., analyzer drift, changes in oxygen uptake rate).

There exists a short-term near-linear relationship between:

• Rate of total ammonia (flow × concentration) entering the system

• The system airflow rate required to achieve an effluent ammonia concentration

This relationship, the specific air, allowed the development of a three-element controller that provides a combination of feedforward, feedback adjustment, and closed-loop aspects within a single controller for dictating system air demand. The function output can handle disturbances and even offset instrument measurement deviation (e.g., air and water flowrates, influent ammonia concentration). This three-element controller creates optimal stability of the system for present conditions and is the basis for the TAAC algorithm.

The three-element controller leverages a static relationship that has been predetermined to be the feedforward aspect of the controller, but this efficiency factor, the specific air analytic, will vary over time for WRRF secondary treatment. Internal system changes that affect the underlying process mechanisms alter the specific air such that the three-element controller requires a large feedback adjustment factor (K) to maintain control through all conditions. The near-linearity of the specific air relationship still exists through these variations, but the controller could benefit from having the ability to compensate for these mechanistic property fluctuations that change the slope.

The feedback adjustment, ⁠, reacts to system disturbances and alters the output, ⁠, to create continuous setpoint achievement of the effluent ammonia concentration. This results in a recent historical dataset that can be considered representative of the optimal slope for current system conditions, as well as predictive of the near future. This historical data can thus be used to generate the slope of the function, ⁠, given that the encompassing function will correct over time to always represent the current process-specific air relationship. Due to this fact, the dynamic model of the specific air is continually valid, or at least continually approaching the present efficiency due to feedback correction. Over time, the calculated value obtained is representative of the present operational conditions; this is a common shortfall of many controllers that are tuned at a momentary operating point (Åström & Hägglund 1995).

During the review of the algorithm, there were some general questions: How much time should be included from the historical data to create the slope function? How drastic are the system disturbances to be handled by the constant of the linear equation (i.e., by how much should the feedback adjustment be allowed to alter the feedforward estimate)?

This section contains the definitions and equations used to create the TAAC algorithm. The statistical rule sets define the value scopes possible for the internal variables of the controller function for any WRRF-activated sludge secondary treatment process aeration system. For this algorithm, time is measured relative to the initiation of the TAAC calculations for control of a system and can be in various increments based on the frequency of process data availability.

Definition: Train

• - In water resource recovery, a train has general use for any series of treatment processes or zones that remove organic matter and nutrients from the substrate. Each process or zone is designed to perform a specific function and to improve the water quality in some way before the discharge is sent back into the environment. In terms of the secondary treatment system being discussed herein, a train comprises of zones that include methods such as anaerobic, anoxic, and aerated reaction methods, which each provide removal of specific components contained within the substrate.

  

  

Definition: Hydraulic retention time (HRT)

• - The time period that a substrate stays within a defined physical region of interest. The HRT between two points is calculated by dividing the volume of each individual process being traversed by the flowrate of the substrate through that area, and then summing all physical transference sections traversed to find the total time period. Examples of substrate transference mechanisms: section of pipe, tank, reactor, lagoon, or any other means of containing or transporting the substrate. This value can be an average for the given physical region found for a time length longer than the HRT of the region. It is even possible for the volume of the region of interest to vary within a train based on variations of the operational purpose and equipment states for each zone within the train.

  

Definition: Train flowrate

• - The total flow of substrate to the initial treatment region of the train. This is often a sum of two flows, the influent supply flowrate and an effluent return flowrate. A flow must enter the train upstream of an aerated zone to be included in the flowrate for that zone. For simplicity, the equations set forth will assume a design such that the return flowrate will enter upstream of all aerated zones.

Definition: TAAC trains – HRT

• - The TAAC train HRT is the transversed region between the physical process point, which the algorithm uses to determine the value of the calculated influent total ammonia and the physical midpoint of all aerated zones of the train. This value can be additionally calculated for each zone within the train, and determination for the individual zone values is recommended in the case of a train where the aerated zones are separated by other nonaerated zones. The equations in this section assume that the influent ammonia concentration measurement is made at a physical process point located upstream of all trains to allow for the overall system to be calculated from a single input value. For the sake of simplicity, the derivation of the following formulas is only for the midpoint for all aerated zones in the train.

Definition: Specific air interval (SAI) time

• - The time period used to find the recent historical specific air of the train.

Note: can be offset from the current time for delays required for data transfer and quality assurance

Ammonia-based aeration control strategies are gaining in popularity due to the improvement of operational treatment performance and a stark reduction of the maintenance and energy costs to facilities. The overall benefits are reduced though when considering the level of knowledge required to implement the strategy, tune the system controller, the frequency in which tuning adjustments are required due to environmental and/or operational changes, and the additional cost for the equipment required. The use of DO probes for control in any ABAC approach comes with the following negative impacts: (1) a significant rise in cost due to the number of field devices, (2) reduced efficiency and performance from probe fouling, and (3) increased maintenance requirements for the process area.

The goal of any ABAC solution is to provide only the air necessary to maintain a stable ammonia discharge from the secondary treatment process, which the controller detailed within this article will optimize by delivering the minimum amount of air at the appropriate time to achieve the desired setpoint. In addition, a controller that does not use DO as a control variable is preferred, as this measurement of process conditions should only be used for deterring overaeration and system air distribution optimization. These items can both be compensated for during the use of the TAAC method by establishing proper algorithm boundaries and through routine DO analysis of the process zones, versus the current method of real-time monitoring. This reduces the amount of instrumentation that is required for current control methods.

There was a pilot test of the algorithm conducted at the Robert W. Hite (RWH) treatment facility, which was a success. The solution was able to maintain a stable effluent ammonia concentration over the multiyear testing period of a single train implementation. The RWH facility runs two distinct liquid process treatment trains that enable major operational shifts to adjust daily facility loading, either for balancing and/or maintenance. These shifts in loading created large changes to influence the flowrate and substrate characteristics for the pilot system, which showed the controller responsiveness. TAAC outperformed all other methods and maintained an effluent ammonia concentration within 0.5 mg/L, with deviations that exceeded these bounds either due to the previously mentioned large swings that would alter the system dynamics or from impacts of other system issues (Figure 5). Future work will be focused on exception handling for these scenarios.

There may be additional components of the algorithm needed to handle system variations and operational conditions under which this solution has not yet been tested and does not presently address, such as biological phosphorus removal. Any subsidiary variables to be considered will simply create an upper or lower boundary of the function output, as is the case with DO for overaeration prevention for energy savings. It is possible that if phosphorus content continues to become a focus area for WRRF discharge permits, then this controller might change or grow, but if this would change the primary variable of concern, it would alter the system paradigm. There is also investigation and research needed into the long-term effects of low levels of DO on BNR systems, but there is little evidence from previous case studies to believe that it will result in significant issues (Ingildsen et al. 2002; Rieger et al. 2012b).

The remainder of this section is a summary of the requirements for operating a facility with the TAAC method. At minimum, the algorithm requires the ability to calculate the system input variables and to monitor the controlled output variable. A converged location for both upstream and downstream of a secondary treatment area allows the option of a single data source for each location to determine the process input and controlled values. This holds true even if the influent process values used in the algorithm are generated using historical data in place of a field device, though this method will provide reduced performance and lessen the analysis capabilities of the internal relationships within the system. This same approach of using historical data can also be used for the air flowrates of each zone for individual trains as a fail-safe.

The functions defined in the appendix show that even in the case of multiple trains within the system, the controller can still function with this same single input and generate a full system output that can then be subdivided among the respective trains. In these use cases, locating at least two effluent probes, within different trains, allows for individual train performance to be assessed during operation. This creates the ability to determine one train's process efficiency, specific air, while a second probe is able to provide feedback data for the remaining operational trains with learned specific air analyses.

A reliable method for ensuring that an accurate effluent ammonia concentration is provided to the algorithm is highly recommended. As few as three probes in the effluent of the secondary process area would be able to provide robust real-time verification. The use of DO probes near train mid- and endpoint(s) can help to detect overaeration from heavy loading or ammonia probe measurement issues and provide upper boundary control of the air flowrate. If a facility plans to operate at the setpoint boundary condition (i.e., ammonia effluent of 0.0 mg/L), it is recommended that DO be incorporated into the overall controller implementation to improve performance and reliability. In these cases, the DO measurement at the same locations mentioned will be able to provide upper boundary control of the TAAC function output.

With the use of the established three-element PID controller design as the basis of the control algorithm, coupled with the generation of a function of the specific air analytic of the secondary treatment using feedback linearization, it is possible to infer that TAAC is the optimal solution for control of a stable effluent ammonia concentration. Given that the mechanistic properties of the system do not change quickly, primarily the nitrifier mass, a window of time of several days can be used for calculating the specific air analytic for the algorithm. This also helps prevent device failure or data quality issues from impacting the controller performance. Even reducing the wastage rate to increase SRT will result in a slow change in nitrifier mass over a period of days and weeks, and the impact on nitrification will be slow (Rieger et al. 2012a).

Finally, if it is considered that the main objective of an aeration controller is to maintain the stability of both the effluent ammonia concentration and total system air demand, then this method will provide the best results with the least cost impacts and overall effort. With all other solutions for ammonia-based aeration control being considered moot through this innovative approach, it is concluded that TAAC will eventually supersede ABAC.

A special thanks to the members of the project team from my first ABAC implementation and paper (W. Anderson, M. Alvis, I. Myers, K. Brischke, and L. Rieger) and to the late Fredrick M. Hayes (d. 2017); together they taught me a great deal about WRRF secondary treatment systems, and that process control requires a team of experts. I would like to mention Craig B. and Jim N.; through their mentorship and efforts I gained, and hopefully retained, a portion of their wisdom and knowledge. The anonymous comments of three reviewers greatly contributed to the framing of the content. This solution has been piloted at Metro Water Recovery, in the south secondary of the Robert W. Hite treatment facility in Denver, CO, and was initiated by the author through the Idea at Work (IAW) employee rewards initiative; a program intended to encourage a culture of innovation. The author would like to thank the personnel of the RWH Operations department, who push for solutions that elevate the work to be done, with the open mindedness and courage to test them, and to all the essential WRRF employees, individuals that daily provide an immeasurable service to their communities and local environment.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.