Abstract
The biological treatment process (BTP) is responsible for removing chemical oxygen demand (COD) and ammonia using microorganisms present in wastewater. The BTP consumes large quantities of energy due to the transfer of oxygen using air pumps/blowers. Energy consumption in the BTP is due to low solubility of oxygen, which results in low aeration efficiency (AE). The aim of the study was to develop an AE model that can be used to monitor the performance of the BTP. Multilayer perceptron artificial neural network (MLP ANN) algorithm was used to model AE in the BTP. The performance of the AE model was evaluated using R2, mean square error (MSE), and root mean square error (RMSE). Sensitivity analysis was performed on the AE model to determine variables that drive AE. The results of the study showed that MLP ANN algorithm was able to model AE. R2, MSE, and RMSE results were 0.939, 0.0025, and 0.05, respectively, during testing phase. Sensitivity analysis results showed that temperature (34.6%), COD (21%), airflow rate (19.1%), and OTR/KLa (15.7%) drive AE. At high temperatures, the viscosity of wastewater is low which enables oxygen to penetrate the wastewater, resulting in high AE. The AE model can be used to predict the performance of the BTP, which will assist in minimizing energy consumption.
HIGHLIGHTS
Aeration efficiency increases with an increase in temperature.
Aeration efficiency decreases with an increase in airflow rate.
Temperature and airflow rate are the biggest driver of energy consumption.
Temperature, COD, and airflow rate control aeration efficiency in the biological treatment process.
High temperatures and airflow rates improve volumetric mass transfer coefficient.
Graphical Abstract
INTRODUCTION
Wastewater treatment plants (WWTPs) are used all over the world including South Africa (SA). WWTPs are responsible for purifying domestic and industrial wastewater to an acceptable effluent quality before being discharged into the environment (Żyłka et al. 2021). The biological treatment process is responsible for decomposing organic matter and oxidising inorganic matter using microorganisms present in wastewater (Kanaujiya et al. 2019). The biological treatment process consumes large quantities of energy due to the transfer of oxygen using air pumps/blowers for respiration and survival of microorganisms (Wang et al. 2016; Christoforidou et al. 2020; Siatou et al. 2020). For example, Wang et al. (2016) reported that the biological energy consumption in Germany, United States of America (USA), and SA were 0.296 kWh/m³, 0.37 kWh/m³, and 0.3–0.61 kWh/m³, respectively. The energy consumption reported in SA, USA, and Germany accounts for 0.55%, 0.6%, and 0.7%, of the annual electricity consumption in their respective countries. In Spain and Saudi Arabia, biological treatment processes consume 0.8 kWh/m³ and 1.6 kWh/m³, respectively (Christoforidou et al. 2020). This energy consumption is high and requires attention in order to achieve low energy consumption.
In order to achieve a high AE, high OTR/KLa must be maintained so that oxygen is supplied in the biological treatment process efficiently. To achieve high OTR/KLa, research scholars have resorted to high airflow rates in order to penetrate the thickness of the wastewater caused by impurities that provide the resistance of oxygen molecules to dissolve in wastewater (Vogelaar et al. 2000; Limpaiboon 2013; Shukla & Goel 2018; Zhou et al. 2018; Warunyuwong & Imai 2020). Utilizing high airflow rates will increase the air velocity distribution which increases the shear force in the biological treatment process. An increase in the shear force will breakdown the thickness of the wastewater making it possible for oxygen molecules to penetrate and dissolve (Ren et al. 2018). According to Zhou et al. (2018), an airflow rate of 0.5 vvm (volume of air per volume of liquid per minute) produced 13.21 kg O2/h OTR/KLa, while an airflow rate of 2 vvm produced 22.43 kg O2/h OTR/KLa. Similarly, Vogelaar et al. (2000) reported that an airflow rate of 0.15 vvm produced 22 kg O2/h OTR/KLa, while an airflow rate of 0.56 vvm produced 61 kg O2/h OTR/KLa. It is clear from the literature that high airflow rates will result in high OTR/KLa. The challenge with increasing airflow rate to obtain high OTR/KLa is that power consumption also increases, resulting in a low AE (Baylar & Bagatur 2000; Pittoors et al. 2014; Al Ba'ba'a & Amano 2017; Collivignarelli et al. 2019; Du et al. 2020; Kan et al. 2020). According to Collivignarelli et al. (2019), an airflow rate of 100 m³/h produced 0.98 kg O2/kWh AE, while an airflow rate of 150 m³/h produced 0.85 kg O2/kWh AE. This is a decrease of 13.3% in AE due to an increase of 33.3% in airflow rate. Similarly, Al Ba'ba'a & Amano (2017) reported that an airflow rate of 0.125 L/min produced 2.8 kg O2/kWh AE, while an airflow rate of 0.05 L/min produced 5.2 kg O2/kWh AE. This was a decrease of 46.2% in AE due to an increase of 60% in airflow rate. Although high airflow rates improve OTR/KLa, they do not benefit AE in the biological treatment process because when high airflow rates are applied, high power consumption are utilized by the air pump/blower. This will result in high energy consumption.
High temperatures, on the other hand, have been reported to improve OTR/KLa, which results in an improvement of AE (Krahe et al. 1996; Vogelaar et al. 2000; Chalasani & Sun 2007; Yang & Park 2012; Pappenreiter et al. 2019). This is because at high temperatures, the viscosity of the wastewater decreases, which reduces the thickness of the wastewater and allows oxygen molecules to penetrate and dissolve more rapidly, resulting in high AE (Lewis & Whitman 1924). According to Vogelaar et al. (2000), a temperature of 20 °C produced OTR/KLa of 17.4 kg O2 h − 1, and a temperature of 55 °C produced OTR/KLa of 37.4 kg O2 h − 1. This is an increase of 53.5% in OTR/KLa due to an increase of 63.6% in temperature. Similarly, Yang & Park (2012) reported that a temperature of 13 °C produced OTR/KLa of 7 kg O2/h while a temperature of 23.5 °C produced OTR/KLa of 7.8 kg O2/h. This was an increase of 10.3% in OTR/KLa due to an increase of 44.7% in temperature. Therrien et al. (2019) reported that a temperature of 10 °C produced OTR/KLa of 0.0324 kg O2/h, and a temperature of 30 °C produced OTR/KLa of 0.0376 kg O2/h. This was a 13.8% gain in OTR/KLa due to a 66.7% increase in temperature. The increase of OTR/KLa due to an increase in temperature will result in an increase of AE because no power consumption was increased. An increase in AE will lead to a reducing in energy consumption in the biological treatment process, because low airflow rates can be utilized with high temperatures to achieve a high OTR/KLa delivery. The challenge is that the normal operating temperature of wastewater is usually in the range between 18 °C and 22.5 °C (Golzar et al. 2020; Tejaswini et al. 2020). This low temperature of wastewater will result in a low OTR/KLa, which will result in low AE due to high viscosity of the wastewater. Therefore, it is essential to develop a model that can be used to monitor AE in order to achieve low energy consumption in the biological treatment process. The aim of this study was to develop an AE model that can be used to monitor the performance of the biological treatment process. The model will incorporate the effects of temperature variations and airflow rate variations since they drive the efficiency of the biological treatment process. This will assist plant managers/operators in achieving high AE while utilizing low airflow rates that reduce energy consumption in the biological treatment process.
MATERIALS AND METHODS
Laboratory experiments
Components of the aeration unit
Components . | Type . | Remarks . |
---|---|---|
Aeration tank | Circular aeration tank (acrylic material) | 10–15 mm thick acrylic material. Dimensions 570 × 300 mm. Total volume is 40 L. Working volume is 30 L |
Dissolved oxygen meter and probe | Hanna HI98196 multi-parameter water proof meter | Measures DO, pH, and ORP |
Thermostat | Local manufacturer | Controls temperature in the range 0 °C–40 °C |
Air pump | Waterfall Resun LP100 | Can supply air between 0 and 140 L/min |
Airflow meter | MF5712 200 L/min digital gas air nitrogen oxygen mass flow meter | Can measure airflow rate between 0 and 200 L/min |
Air stone disc bubble diffuser | Growneer micro pore bubble diffuser | Diffuser is 20 cm in diameter |
Digital wattmeter | Geewiz (kilowatt) | Measures power between 0 and 3,600 watts |
Components . | Type . | Remarks . |
---|---|---|
Aeration tank | Circular aeration tank (acrylic material) | 10–15 mm thick acrylic material. Dimensions 570 × 300 mm. Total volume is 40 L. Working volume is 30 L |
Dissolved oxygen meter and probe | Hanna HI98196 multi-parameter water proof meter | Measures DO, pH, and ORP |
Thermostat | Local manufacturer | Controls temperature in the range 0 °C–40 °C |
Air pump | Waterfall Resun LP100 | Can supply air between 0 and 140 L/min |
Airflow meter | MF5712 200 L/min digital gas air nitrogen oxygen mass flow meter | Can measure airflow rate between 0 and 200 L/min |
Air stone disc bubble diffuser | Growneer micro pore bubble diffuser | Diffuser is 20 cm in diameter |
Digital wattmeter | Geewiz (kilowatt) | Measures power between 0 and 3,600 watts |
Wastewater collection
Wastewater samples were collected from the Daspoort WWTP bioreactor number two each day. Two samples were collected, one in the morning and one in the afternoon. The collection method (morning and afternoon) was implemented to prevent any biological activity from taking place during the storage process. The APHA (2012) method was used for the collection of wastewater from the Daspoort WWTP. Non sterile powder free nitrile gloves were used to handle all equipment. A 25 L jerry can was used to collect the raw wastewater. An additional 2 L of activated sludge was collected from the return activated sludge. Wastewater samples were collected and aerated within 2 h after collection. After collection, the wastewater was stored in ice bucket at a temperature of 4 °C to make sure that no biological activities take place during transportation to the Tshwane University of Technology, Faculty of Engineering and the Built Environment, Department of Civil Engineering.
Analysis
Chemical oxygen demand (COD) was measured and analysed. The method used for analysing COD was the Standard Methods for the Examination of Water and Wastewater (APHA 2012). COD was analysed using closed reflux colorimetric (APHA 2012 method 5220). Spectrophotometer (DR3900), digestion vessels, block heater, microburet, and ampule sealer were the equipment used for the analysis.
Aeration process
Aeration process was conducted twice a day (two samples collected). Two wastewater samples were aerated per day and each sample was aerated for 4 h (hydraulic retention time at Daspoort WWTP). This means that two different temperatures (35 °C and 32.5 °C) were applied during the aeration process at one of the airflow rates (5 L/min) per day. The same process was repeated the following day applying two different temperatures (30 °C and 27.5 °C) and the same airflow rate (5 L/min). The process was then repeated on all temperatures at different airflow rates. A thermostat (Table 1) was used to increase (heat up) the temperature of wastewater from field temperature (18–22.5 °C) to the desired operating temperature. Ice cubes were used to decrease (reduce) the temperature of wastewater from field temperature to the desired operating temperature. Three samples were collected and analysed during each aeration process:
Initial concentration immediately after collection or before aeration process.
2 h concentration.
4 h concentration.
Disposal
After the completion of the aeration process, wastewater was disposed by flushing at the laboratory toilet. The disposal method allows the wastewater to go back to the WWTP.
Determination of oxygen uptake rate, oxygen transfer rate/volumetric mass transfer coefficient, and aeration efficiency
The dynamic method as described by Garcia-Ochoa & Gomez (2009) was used to determine the oxygen uptake rate (OUR) and OTR/KLa.
Oxygen uptake rate
Five steps were followed to determine OUR (Garcia-Ochoa & Gomez 2009):
- i.
Step one – the airflow into the aeration unit was turned off for 10 min, which was enough to prevent microorganisms from being starved of oxygen.
- ii.
Step two – the decrease in DO concentration was monitored during the non-aeration period using a DO meter and probe.
- iii.
Step three – the decrease in DO was plotted against time using Microsoft Excel.
- iv.
Step four – a linear regression line was used to determine the rate of change in DO concentration with time. The rate of change in DO concentration was regarded as the OUR.
- v.Step five – Equation (2) defines the rate of change in DO concentration with time.
Oxygen transfer rate/volumetric mass transfer coefficient
Seven steps were followed to determine OTR/KLa (Garcia-Ochoa & Gomez 2009).
- i.
Step one – turn on the air supply in the aeration unit.
- ii.
Step two – record the DO concentration at different time intervals.
- iii.
Step three – measure the average DO concentration saturation in the biological aeration unit.
- iv.Step four – determine the natural logarithm of the difference between DO concentration at different time intervals and average DO concentration obtained at infinite time using Equation (3).
- v.Step five – determine the difference between time intervals using Equation (4).
- vi.
- vii.
Step seven – produce a linear regression line. The slope of the linear regression line is equal to the OTR/KLa.
Aeration efficiency
Three steps were followed to determine AE (Garcia-Ochoa & Gomez 2009).
- i.
Step one – determine OTR/KLa in the aeration unit. OTR/KLa will be determined as described in section 2.6.2.
- ii.
Step two – determine the power used during the aeration process. Power input was determined as described in Table 1, using the digital wattmeter.
- iii.
Step three – determine the AE. AE was determined by the quotient of OTR/KLa and power input as described in Equation (1).
Application of artificial neural network




The first step was to select input variables from the data collected during laboratory experiments.
The second step was to split the data into 70% training, 15% validation, and 15% testing.
- The third step was to transform data into input variables using the normalization technique defined by Equation (6).where
is the scaled sample data point, x is the sample data point,
is the minimum value in the training dataset,
is the maximum value on the training dataset.
The fourth step was to select the MLP ANN architecture with one hidden layer and four neurons.
The fifth step was to initialize the weights and bias between values of zero and one.
The seventh step was to use the supervised gradient descent backpropagation learning algorithm to correct the weights and bias.
The ninth step was to perform sensitivity analysis on the AE model. A one factor at a time method was used. Each decision variable was individually adjusted by ± 10%, ± 20%, and ± 50%, while keeping the others unchanged. In addition, decision variables were checked for their direct, inverse, major, or minor changes on AE.
RESULTS AND DISCUSSION
Laboratory results
Pearson correlation coefficient (2 tailed) between aeration efficiency and the parameters of interest
Correlation . | |||||||
---|---|---|---|---|---|---|---|
. | Temperature . | OUR . | OTR/KLa . | COD . | Airflow rate . | AE . | |
Temperature | Pearson Correlation | 1 | |||||
OUR | Pearson Correlation | 0.271** | 1 | ||||
OTR/KLa | Pearson Correlation | 0.305** | 0.747** | 1 | |||
COD | Pearson Correlation | 0.123** | −0.695** | −0.579** | 1 | ||
Airflow rate | Pearson Correlation | 0.000 | 0.097** | 0.229** | 0.008 | 1 | |
AE | Pearson Correlation | 0.575** | 0.206** | 0.542** | −0.043** | −0.061** | 1 |
Correlation . | |||||||
---|---|---|---|---|---|---|---|
. | Temperature . | OUR . | OTR/KLa . | COD . | Airflow rate . | AE . | |
Temperature | Pearson Correlation | 1 | |||||
OUR | Pearson Correlation | 0.271** | 1 | ||||
OTR/KLa | Pearson Correlation | 0.305** | 0.747** | 1 | |||
COD | Pearson Correlation | 0.123** | −0.695** | −0.579** | 1 | ||
Airflow rate | Pearson Correlation | 0.000 | 0.097** | 0.229** | 0.008 | 1 | |
AE | Pearson Correlation | 0.575** | 0.206** | 0.542** | −0.043** | −0.061** | 1 |
**Correlation is significant at the 0.01 level (2-tailed).
*Correlation is significant at the 0.05 level (2-tailed).
Relationship between aeration efficiency and COD (a), airflow rate (b), temperature (c), OUR (d), and OTR (e).
Relationship between aeration efficiency and COD (a), airflow rate (b), temperature (c), OUR (d), and OTR (e).
Airflow rate (−0.061**) produced a negative correlation relation to AE. At high airflow rates, air bubbles tend to move faster, resulting in a shorter retention time of air bubbles that contain oxygen in the biological treatment process. The shorter retention time means that less oxygen molecules will be absorbed in the wastewater, yet high power is being consumed, resulting in low AE (Ashley et al. 2008; Al Ba'ba'a et al. 2014; Collivignarelli et al. 2019). According to Ashley et al. (2008), an airflow rate of 10 L/min produced 270 kg O2/kWh AE, while an airflow rate of 40 L/min produced 180 kg O2/kWh AE. Similarly, Al Ba'ba'a et al. (2014) reported that an airflow rate of 14.4 L/min produced 64.05 kg O2/kWh AE while an airflow rate of 84 L/min produced 2.94 kg O2/kWh AE. Figure 4(b) shows a similar relationship between AE and airflow rate. This shows that high airflow rates are not beneficial in the biological treatment process. Temperature (0.575**) produced the highest positive correlation relation to AE. This was because temperature controls the viscosity of wastewater, and at high temperatures the viscosity of wastewater is low, which increases the transfer of oxygen molecules in wastewater. When oxygen transfer is high in wastewater due to low viscosity, AE will increase because less power from air pumps/blowers is required to achieve high oxygen solubility (Therrien et al. 2019). This can also be justified by the relationship between temperature and OTR/KLa, which produced a positive correlation of 0.305**. This relationship shows that when the viscosity of wastewater is low due to high temperatures, oxygen transfer increases due to a thin liquid film that allows for high oxygen transfer. Therrien et al. (2019) reported that a temperature of 10 °C produced 0.197 kg O2/kWh AE and a temperature of 30 °C produced 0.222 kg O2/kWh AE. This resulted in a 11.3% increase due to an increase in temperature of wastewater. The relationship between AE and temperature can be observed in Figure 4(c).
OUR (0.206**) produced the third highest positive correlation relation to AE. In the biological treatment process, OTR/KLa is equal OUR when airflow rate is maintained at a constant because the rate of change in dissolved oxygen concentration remains the same. In addition, high OUR indicates high oxygen consumption by microorganisms, which suggests that oxygen was available in abundance due to high OTR/KLa, resulting in high AE. This can be justified by the relationship between OTR/KLa and OUR, which produced a positive correlation of 0.747** in the biological treatment process. The graphical relationship between AE and OUR can be seen in Figure 4(d). OTR/KLa (0.542**) produced the second highest positive correlation relation to AE. When OTR/KLa increases in the biological treatment process, it indicates that more mass of oxygen is being transferred into the wastewater using less power consumption, hence AE increases. In addition, OTR/KLa is directly proportional to AE as defined by Equation (1), hence they produce a positive relationship. The relationship between AE and OTR/KLa can be observed in Figure 4(e).
Modelling performance results
MLP ANN energy consumption model performance
. | Training . | Validation . | Testing . |
---|---|---|---|
R2 | 0.963 | 0.963 | 0.939 |
MSE | 0.0012 | 0.00075 | 0.0025 |
RMSE | 0.0353 | 0.0274 | 0.05 |
. | Training . | Validation . | Testing . |
---|---|---|---|
R2 | 0.963 | 0.963 | 0.939 |
MSE | 0.0012 | 0.00075 | 0.0025 |
RMSE | 0.0353 | 0.0274 | 0.05 |
The blue, green, and red lines in Figure 8 indicate training, validation, and testing phases, respectively. The best line in Figure 8 indicates the lowest MSE and number of iterations during the validation phase of the model. Because the MSE was closer to zero during training, validation, and testing phases, it shows that the difference between the predicted and observed data was small. The results obtained in this study showed that MLP ANN algorithm captured the AE data accurately in the biological treatment process.
The RMSE during training, validation, and testing phase results were 0.0353, 0.0274, and 0.05, respectively, as shown in Table 3. The optimum RMSE was determined during the validation phase, which justifies the optimum MSE obtained during the validation phase. The difference between the highest and lowest RMSE was 29.4%, which was a significant difference. Overall, MLP ANN AE model performed well in all learning phases. Results obtained in this study were superior compared with the results obtained by other scholars (Hamada et al. 2018; Newhart et al. 2020; Aalami et al. 2021). Newhart et al. (2020) applied ANN algorithm to model acid disinfection performance in a WWTP and the model produced RMSE of 6.95. Similarly, Aalami et al. (2021) applied ANN algorithm to model the performance of a WWTP and the model produced RMSE of 3.93. Hamada et al. (2018) applied ANN algorithm to model the performance of a WWTP and the model produced RMSE of 29.69.
R2 during training phase (a), validation phase (b), and testing phase (c).
Similar results were obtained by the following scholars (Kundu et al. 2013; Tumer & Edebali 2015; Hamada et al. 2018; Bekkari & Zeddouri 2019; Ofman & Struk-Sokołowska 2019; Oulebsir et al. 2020; Alsulaili & Refaie 2021). Alsulaili & Refaie (2021) applied ANN algorithm to model five day biological oxygen demand in wastewater treatment and the model produced R2 value of 0.754. Bekkari & Zeddouri (2019) applied ANN algorithm to predict effluent COD in wastewater and the model produced R2 of 88%. Ofman & Struk-Sokołowska (2019) applied ANN algorithm in modelling nitrogen forms in wastewater and the model produced R2 of 0.99. Oulebsir et al. (2020) modelled the activated sludge energy consumption. R2 values obtained were 94.41% and 82.42% during training and testing phases. Similarly, Tumer & Edebali (2015) applied ANN algorithm to model the biological treatment process and the model produced R2 of 96%. Kundu et al. (2013) applied ANN algorithm to model the biological removal of organic carbon and nitrogen in wastewater and the model produced R2 value of 96%. Hamada et al. (2018) applied ANN algorithm to model the performance of a WWTP and the model produced R2 of 78%.
Sensitivity analysis
Effect of COD (a), airflow rate (b), temperature (c), OUR (d), and OTR (e) on aeration efficiency.
Effect of COD (a), airflow rate (b), temperature (c), OUR (d), and OTR (e) on aeration efficiency.
CONCLUSION
The aim of the study was to develop an AE model that can be used to monitor the performance of the biological treatment process. The following data collected from the laboratory was used: COD, temperature, airflow rate, OUR, and OTR/KLa. Low concentration of COD showed high AE. High AE was measured at low airflow rate of 15 L/min. High temperatures (35 °C) produced the highest AE of 0.621 kg O2/kWh. The highest AE (0.621 kg O2/kWh) was measured at moderate OUR of 0.2567 mg/l. AE increased with an increase in OTR/KLa in the biological treatment process. MLP ANN algorithm was used to model the AE data. MLP ANN algorithm was successful in modelling the AE data. R2 values were 0.963, 0.963, and 0.939 for training, validation, and testing phases respectively. MSE values were 0.0012, 0.00075, and 0.0025 for the training, validation, and testing phases, respectively. RMSE values were 0.0353, 0.0274, and 0.05 for the training, validation, and testing phases, respectively. Sensitivity analysis showed that temperature (34.6%) and COD (21%) were the biggest drivers of AE in the biological treatment process. The AE model can be utilized by plant managers/operaters to improve AE which ultimately improves energy consumption in the biological treatment process.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.