In recent years, China has begun to reap the benefits of installing and operating automatic process control systems in wastewater treatment plants. Multiple case studies have demonstrated that an advanced model based control system may not only assist wastewater treatment plants in meeting more stringent effluent permits, but also lead to improvements in process stability and energy savings. To achieve these benefits and ensure the plant is operating at peak efficiency, Bailonggang Wastewater Treatment Plant (WWTP), located in Shanghai, China, has opted to employ an advanced model based aeration control system in its most recent expansion which added 212 MGD of capacity to the existing capacity of 528 MGD. This paper presents a full case study of the process of designing and commissioning an aeration system based upon variable oxygen uptake rate and discusses the typical issues which arise while implementing such a system in such a large scale application.
Shanghai Bailonggang WWTP, China – Plant Profile
Shanghai Bailonggang WWTP is in the process of being designed and built in a series of phases. Currently, this plant is responsible for serving an area of 105 mi2 (272 km2) and a population of 3.56 million. Total flow to the plant is expected to reach 951 MGD by 2020. Phase I of the plant has a design capacity of 528 MGD, which was reached in 2008.
表1 上海白龍港污水處理廠II期擴建工程設(shè)計進、出水水質(zhì)
項目 | COD （mg/L） | BOD5 （mg/L） | SS （mg/L） | TN （mg/L） | NH3-N （mg/L） | TP （mg/L） |
進水 | 350 | 150 | 170 | 40 | 30 | 5.0 |
出水 | 60 | 20 | 20 | 20 | 15 | 1.0 |
Planning for a Phase II expansion project began in early 2010 with the expectation for it to be fully operational with a capacity of 212 MGD in April 2013. The secondary treatment process for Phase II consists of eight identical trains in an anaerobic-anoxic-oxic (AAO) configuration. A diagram of one train can be found in Figure 1 which indicates DO control zones, air pipes and the installation positions for analyzers and other equipment.
圖1 Layout of One Aeration Train and Installation Positions for Instruments
METHODOLOGY
Control Theory
Wastewater treatment plants (WWTPs) do not typically run at design conditions throughout the lifetime of the plant, conditions often varying widely on a daily basis. This dynamic influent characteristic provides an ongoing opportunity for system optimization. Implementing an aeration control system using dynamic DO set-points which responds to these continuously variable influent conditions provides overall lower energy consumption by reducing the aeration demand when appropriate.
The purpose of calculating and tracking dynamic DO set-point is to meet effluent ammonia targets while minimizing the amount of oxygen supplied to the process. Minimizing the supplied oxygen reduces the loading on the air blowers, resulting in significant energy and monetary savings.
Residual DO in each aerobic stage of the process can be used to control the specific nitrification rate of the biomass. Increasing the DO concentration increases the nitrification rate by increasing the availability of oxygen to the autotrophic biomass. The increase in the nitrification rate results in a decrease in the effluent ammonia concentration. The converse is also true; lowering the DO concentration will increase effluent ammonia.
Feed-Forward Process Controller Description
The Feed-Forward Process Controller is a combination of two control systems. One is a model-based process controller and the other is an aeration control system. Both systems operates to assist the facility to control the BNR process properly by adjusting sensitive system parameters such as internal recycle flow, return activated sludge and waste stream flows, DO set-points, individual zone airflow set-points, valve position set-points, and blower airflow set-points.. Figure 3 is a diagram that visually explains the basic operation of the feed-forward process controller.
Model-based Process Controller
The model-based process controller is designed to use a feed-forward control algorithm to calculate the proper DO set points for each control zone in the aerobic stages. By monitoring the influent flow rate and concentration and receiving an effluent target which operators can set anytime on the interface, this model-based system uses an optimization algorithm to determine the effect different DO set points will have on the effluent ammonia concentration and ensures the DO set-points system finally calculates are the lowest possible ones to maintain the effluent ammonia targets.
The fundamental on-site parameters for establishing the model are divided into two respects. Those data which is relating to the activated sludge is measured by the experimental equipment called “ABAM” which is used to detect the sludge Oxygen Uptake Rate (OUR), Nitrification Rate (NR), and Denitrification Rate (DNR). The composition of the influent loading is analyzed in the lab to determine some ratio values for the model, such as BOD/COD, TKN/COD, and NH4/TKN.
Feed Forward Controller Hardware and Program
A touch screen industrial computer is used for the installation of the model-based controller program due the heavy calculation load. A front end Human Machine Interface (HMI) was developed using Microsoft Visual Basic for the industrial computer allowing operational modes, set points, and soft coded tuning parameters to be accessed and modified directly and easily. The industrial computer also contains a data logging and plotting module powered via a SQL based data service which runs concurrently with the control software.
The process control engine, where the process model and optimization algorithm calculations are implemented, was developed utilizing Matlab and consists of a specially tuned set of equations based upon the Activated Sludge Model I (ASMI) developed by the International Water Association (IWA).
Results
Figure 6: Actual DO Reading under Manual Control over One Week
Figure 7: Actual DO Reading under Automatic Control over One Week
The difference in energy consumption by the blower system between manual control and automated control was also measured. To account for differences in the influent flow rate between the two controlled periods, the measured energy consumption of the aeration blowers was normalized per 10,000 metric tons of influent flow.
Two months of influent data, shown in Figure 8, from when the plant was under manual control was used to calculate baseline energy consumption. The average energy consumption to treat 10,000 metric tons of influent flow was calculated as 1038.59 kWh while the average influent ammonia was 30.72mg/L.
Figure 8: Energy Consumption baseline for manual control of two months of 2013
Two different time ranges were compared with the manual control basin line. These periods are shown in Figures 9 and 10. The average energy consumptions to treat 10,000 metric tons of influent flow were 835.09 kWh and 827.29 kWh for the two time periods, while the average influent ammonia was 36.99mg/L and 28.89mg/L.
Figure 9: Energy Consumption for system control of two months of 2013
Figure 10: Energy Consumption for system control of one month of 2014
This represents an average savings of 19.9% compared to manual control. Assuming a 0.962 CNY per kilowatt-hour electricity price for summer (from June to September) and a 0.657 CNY per kilowatt-hour electricity price for the rest of the year, the advanced aeration control system is expected to save an average of 267,499 CNY ($42,460) per month.
Discussions
In order to implement and run a Feed-forward process control system in a plant with such a large scale, unique control solutions had to be developed to overcome issues which do not normally occur at smaller scale WWTPs.
First, the blower control was more difficult than typically encountered. The blower system manufacturer has its own dead-band calculation method which is based upon the total number of blowers. The blower system has 6 blowers with four of them expected to be operating and the other two are for back-up. Therefore, the dead-band of the entire blower system was calculated using the sum of four blowers’ maximum airflow rates, which resulted in a dead-band of 4800m3/H (the Phase 2 total airflow rate is typically between around 80,000m3/H – 110,000m3/H). Figure 11 shows the tracking accuracy under this dead-band setting.
Figure 11: Blower Tracking Performance Under 4800m3/H dead-band.
This initial dead-band, however, is too large to achieve optimal control for the system especially considering the system is currently only running two of the four total blowers in operation. In an effort to compromise with the blower manufacturer, a lower dead-band of 2400 m3/H was pushed into the blower controller resulting in more accurate total airflow tracking performance as depicted below in figure 12.
Figure 12: Blower Tracking Performance Under 2400m3/H dead-band.
Secondly, each additional data collection device introduces an additional point of error or failure to a control system which relies on the health of that signal. To accommodate for the large number of airflow meters employed, and the subsequently increased risk of failure that follows, a sophisticated airflow multiplexing/redundancy algorithm was configured which allows for discrepancies in signals generated by both individual zone airflow meters and total airflow meters to be reconciled despite individual unit miscalibration and unpredictable unit failures. The result is a smooth and robust feedback signal which the blower system can utilize despite occasional airflow meter malfunction.
Conclusions
Implementing an advanced aeration control system in Phase 2 of Bailonggang WWTP has not only improved DO control performance and stability with varying influent loading conditions, but it also has resulted in a substantial improvement in energy savings and a reduced cost of operations. It can also be concluded that process optimization techniques including aeration control algorithms scale effectively into especially large facilities such as this one.