Evaluating Model Predictive Control for Aeration System Efficiency and Reliability

Introduction To meet water quality regulations and the needs of a growing service area, Metro Water Recovery (Metro) is partnering with APG-Neuros to install and pilot APG-Neuros' Advanced Aeration Control (AAC) system at the Northern Treatment Plant (NTP). AAC uses state-of-the-art model predictive control (MPC) algorithms to improve controller performance compared to traditional control methods. The goal of piloting this advanced control system was to demonstrate the ability of the AAC to meet valve position, airflow, dissolved oxygen (DO), and effluent ammonia setpoints as determined by the model outputs. This presentation will provide results of this pilot including the ability of the controller to meet the desired setpoints during two phases: (1) the initial commissioning and operation in the spring/summer of 2023 and (2) following the commissioning of an additional bioreactor in the winter/spring 2024. NTP Site and Process Description The NTP is a regional water resource recovery facility located in Brighton, Colorado, operated by Metro since 2016. With an average influent flow of 7.4 million gallons per day (mgd) in 2023, the NTP uses a step-feed, five-stage Bardenpho-type process to achieve a low effluent total nitrogen concentration. The plant has operated two bioreactors, and an additional bioreactor has been brought online to meet an expected 4.5-mgd increase in flows in 2024. Aeration to eight bioreactors is supplied by six APG-Neuros NX350-C100 High Speed Turbo blowers on a common header, which also supplies air to NTP's Post Digestion Solids Treatment (PDST). Historically, the facility has used a Proportional-Integral-Derivative (PID) based DO control with user-defined DO setpoints to control the bioreactors' aeration. With this complex flow system containing 24 modulating valves requiring phosphorus removal, low nitrogen limits, and increasing flows and loads, the PID controller struggles to maintain consistent control. Controller Description The new APG Neuros AAC consists of several cascaded control loops working in series to ensure effluent ammonia discharge targets are achieved using the lowest airflow possible. Ammonia-based aeration control (ABAC) logic varies the DO setpoints based on the ammonia loading. The blower speed(s) are then adjusted to produce the required amount of total airflow. All control loops, including the ABAC loop, use MPC instead of PID logic. An MPC controller consists of two parts, a predictor and an optimizer. Methodology Each aeration zone was calibrated with all eight aeration zones being controlled individually by the AAC to provide the model with historical data on valve reaction time and DO response. From there, the AAC controlled the bioreactor in a two-setpoint manner, with an ammonia setpoint provided for the first pass and a lower setpoint in the second pass. Due to influent loading increases and mechanical aeration restrictions that adversely impacted single-setpoint control, phase 1 testing only used two-setpoint control. During phase 1, Bioreactor 4 was placed into AAC mode from April to August 2023. The performance of Bioreactor 4 was compared against its own historical operation under DO control and against Bioreactor 2. Effluent ammonia target concentration and air demand were the primary performance criteria. Data was analyzed for a 24-hour period on July 29, 2023 at one-minute intervals and compared to data from the DO control strategy used at NTP on February 27, 2023. A comparison of DO, effluent ammonia, valve position, and airflow setpoints was conducted to determine the accuracy and stability of the AAC in maintaining the determined setpoints. During phase 2, Bioreactor 4 will be operated side-by-side with Bioreactor 3 for three months to compare AAC with DO control using a single ammonia setpoint for full bioreactor aeration control. Results and Discussion The initial assessment of the AAC performance showed that the airflow and valve position accurately operated at the determined setpoint in AAC operation (Figure 1). DO setpoint tracking and stability also improved with AAC control compared to DO control (Figure 2). During AAC testing, there was a period of significant DO setpoint fluctuations when the system was able to accurately aerate to maintain the DO setpoint despite the significant change in the AAC-generated DO setpoint. This speaks to the AACs ability to effectively predict DO setpoints and adjust the system's aeration to meet the aeration demands. The airflow and valve operation and setpoints with AAC are smoother (less erratic) and do not change drastically during the day. This is important as constant and significant changes to the valve positions can cause valve wear that can lead to premature failure. For Bioreactor 4 Pass I, the process performance, when comparing influent ammonia and effluent ammonia, was significantly better with AAC as compared to DO control. Overall, the system stability and predictability with AAC significantly improved equipment operation and overall process performance as shown by the Pass I effluent ammonia concentration (Figure 3). Further investigation and evaluation in a Phase 2 evaluation will provide a more comprehensive picture into how the dynamics of the system affect aeration controls and controller performance.