Reducing aeration energy consumption in a large-scale membrane bioreactor: Process simulation and engineering application

Accepted Manuscript Reducing aeration energy consumption in a large-scale membrane bioreactor: Process simulation and engineering application Jianyu Sun, Peng Liang, Xiaoxu Yan, Kuichang Zuo, Kang Xiao, Junlin Xia, Yong Qiu, Qing Wu, Shijia Wu, Xia Huang, Meng Qi, Xianghua Wen PII:

S0043-1354(16)30085-9

DOI:

10.1016/j.watres.2016.02.026

Reference:

WR 11842

To appear in:

Water Research

Received Date: 25 October 2015 Revised Date:

18 January 2016

Accepted Date: 11 February 2016

Please cite this article as: Sun, J., Liang, P., Yan, X., Zuo, K., Xiao, K., Xia, J., Qiu, Y., Wu, Q., Wu, S., Huang, X., Qi, M., Wen, X., Reducing aeration energy consumption in a large-scale membrane bioreactor: Process simulation and engineering application, Water Research (2016), doi: 10.1016/ j.watres.2016.02.026. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT 1

Date: Jan. 18, 2016

2

Resubmitted to: Water Research

3

Reducing aeration energy consumption in a large-scale membrane bioreactor:

5

process simulation and engineering application

6

RI PT

4

Jianyu Suna, Peng Lianga, Xiaoxu Yana, Kuichang Zuoa, Kang Xiaoa,b,**, Junlin Xiaa,

8

Yong Qiua, , Qing Wua, Shijia Wua, Xia Huanga,*, Meng Qia, Xianghua Wena

M AN U

SC

7

9 10

a

11

School of Environment, Tsinghua University, Beijing 100084, China

12

b

13

Beijing 100049, China

14

* Corresponding author, Tel: +86 10 62772324; E-mail: [email protected]

15

** Co-corresponding author, Tel: +86 10 88256452; E-mail: [email protected]

EP

TE D

College of Resources and Environment, University of Chinese Academy of Sciences,

AC C

16

State Key Joint Laboratory of Environment Simulation and Pollution Control,

1

ACCEPTED MANUSCRIPT Abstract

18

Reducing the energy consumption of membrane bioreactors (MBRs) is highly

19

important for their wider application in wastewater treatment engineering. Of

20

particular significance is reducing aeration in aerobic tanks to reduce the overall

21

energy consumption. This study proposed an in situ ammonia-N-based feedback

22

control strategy for aeration in aerobic tanks; this was tested via model simulation and

23

through a large-scale (50,000 m3/d) engineering application. A full-scale MBR model

24

was developed based on the activated sludge model (ASM) and was calibrated to the

25

actual MBR. The aeration control strategy took the form of a two-step cascaded

26

proportion-integration (PI) feedback algorithm. Algorithmic parameters were

27

optimized via model simulation. The strategy achieved real-time adjustment of

28

aeration amounts based on feedback from effluent quality (i.e., ammonia-N). The

29

effectiveness of the strategy was evaluated through both the model platform and the

30

full-scale engineering application. In the former, the aeration flow rate was reduced

31

by 15–20%. In the engineering application, the aeration flow rate was reduced by

32

20%, and overall specific energy consumption correspondingly reduced by 4% to 0.45

33

kWh/m3-effluent, using the present practice of regulating the angle of guide vanes of

34

fixed-frequency blowers. Potential energy savings are expected to be higher for

35

MBRs

36

ammonia-N-based aeration control strategy holds promise for application in full-scale

37

MBRs.

AC C

EP

TE D

M AN U

SC

RI PT

17

with

variable-frequency

blowers.

38 2

This

study

indicated

that

the

ACCEPTED MANUSCRIPT 39

Keywords:

40

Membrane bioreactor, ammonia-N-based aeration control strategy, simulation,

41

engineering application

43

RI PT

42

1 Introduction

Membrane bioreactors (MBRs) have been widely used for wastewater treatment

45

and reclamation engineering over the past decade (Huang et al., 2010; Judd & Judd,

46

2011; Xiao et al., 2014). Compared with conventional activated sludge (CAS)

47

processes, MBRs possesses advantages such as high effluent quality and low sludge

48

production, and provide promising alternatives in areas facing water crises and

49

needing to meet stringent pollutant discharge standards. However, the operational cost

50

of MBRs, especially related to energy consumption, has typically been greater than

51

that of CAS systems (Judd & Judd, 2011). The majority of energy is consumed for

52

aeration, both in biological tanks for pollutant degradation, and in membrane tanks for

53

retarding membrane fouling (Brepols et al., 2010; Yan et al., 2015). The energy

54

consumption of aeration accounts for 70–80% of total consumption of the wastewater

55

treatment process. A total of 40–60% of aeration energy is consumed in biological

56

tanks (Xiao et al., 2014). The optimization of aeration in MBRs is therefore of

57

practical importance to reduce operational costs and increase the competitiveness of

58

MBRs.

AC C

EP

TE D

M AN U

SC

44

59

The optimization of aeration could be achieved via regulation of blowers,

60

including by altering the angle of the guide vanes for fixed-frequency blowers and by 3

ACCEPTED MANUSCRIPT altering the frequency of the alternating current for variable-frequency blowers.

62

Aeration can be controlled manually or automatically, on the basis of process

63

parameters, e.g., mixed liquor suspended solids (MLSS) concentrations (Zha et al.,

64

2006), dissolved oxygen (DO) concentrations (Ingildsen et al., 2002), and

65

oxidation-reduction potential (ORP) (Fatone et al., 2008). Among these, DO-based

66

automatic aeration control has attracted most attention in both simulations and

67

experimental validations aiming to reduce energy consumption (Gabarron et al., 2015;

68

Ingildsen et al., 2002; Pittoors et al., 2014; Wang et al., 2007). DO-based aeration

69

control aims to maintain a stable DO concentration in biological tanks. Energy

70

consumption reduction and better nutrient removal performance can be achieved by

71

optimizing the DO setpoint of the control (Gabarron et al., 2015). However, DO is

72

merely an intermediate variable between aeration and effluent quality. The

73

determination of optimal aeration demand is a trade-off between aeration amount and

74

effluent quality. Moreover, maintaining DO concentrations at a so-called “optimal”

75

level during operations could still lead to real-time fluctuations in effluent quality

76

under dynamic influent conditions, meaning that aeration is never fully optimized.

77

Given the above, a feedback control mechanism that has its starting point in effluent

78

quality (rather than DO) might have greater potential for aeration optimization and

79

hence energy saving.

AC C

EP

TE D

M AN U

SC

RI PT

61

80

Model tools are essential for automatic feedback control, which is expected to

81

enable real-time determination of aeration demand according to effluent quality.

82

Establishment of a quantitative relationship between effluent quality and aeration is of 4

ACCEPTED MANUSCRIPT paramount importance. The relationship should itself be precise and simple enough

84

for calculation that allows real-time control. However, the well-known activated

85

sludge model (ASM), which has been broadly used for describing the wastewater

86

treatment process (Hauduc et al., 2013), consists of multiple differential formulas

87

(Henze et al., 1999). The solving process is complicated, making it difficult to meet

88

real-time control requirements. If a straightforward empirical relationship, which is

89

simple in terms of calculation but matches well with the complete ASM, could be

90

extracted from the latter, this would be of great benefit for real-time feedback control

91

in practical wastewater treatment systems.

M AN U

SC

RI PT

83

When describing the wastewater treatment process of MBRs via the ASM, the

93

following considerations are particularly important. First, membrane rejection

94

prolongs solid retention time (SRT) and lowers the food-to-microorganisms (F/M)

95

ratio in MBR, resulting in significant differences in kinetic parameters between MBR

96

and CAS (Fenu et al., 2010). Second, near-saturated DO concentrations are achieved

97

in membrane tanks due to abundant aeration for membrane scouring (Sun et al., 2014).

98

The DO-enriched mixed liquor of the membrane tank is usually recirculated back to

99

the aerobic zone of the biological tanks, which can render the oxygen balance

101

EP

AC C

100

TE D

92

between MBR and CAS systems quite different (Sun et al., 2015). This study aimed to reduce aeration energy consumption in a large-scale MBR

102

via process simulation and engineering application. First, an MBR-adapted ASM

103

model was calibrated to achieve an acceptable simulation of the full-scale MBR.

104

Second, an effluent-quality-based feedback control strategy for aeration in biological 5

ACCEPTED MANUSCRIPT tanks was established, comprising a two-step cascaded proportion-integration (PI)

106

control process; this involved establishing a simple mathematical relationship

107

between aeration and effluent quality, which was also accordance with the ASM. The

108

strategy calculated a dynamic DO setpoint according to in situ ammonia-N

109

concentrations, then calculating a dynamic aeration flow rate for adjusting the blowers.

110

The algorithmic parameters were optimized via tentative model simulation. Third, this

111

strategy with optimized parameters was applied to practical operation of a large-scale

112

MBR, to validate the feasibility of the control strategy. Energy saving potential was

113

assessed via model simulation, and actual effectiveness was further examined via

114

practical application, by comparing energy levels and in situ ammonia-N

115

concentrations before and after application of the aeration control strategy.

M AN U

SC

RI PT

105

TE D

116 117

2 Material and methods

118

2.1 Full-scale MBR

The full-scale MBR was installed in a municipal wastewater treatment plant

120

(WWTP). The WWTP, with a capacity of 50,000 m3/d, was located in Wuxi (30°36’ N,

121

120°19’ E, Jiangsu Province, south of China) and treated a cocktail of domestic and

122

industrial wastewaters. The MBR process was operated at hydraulic and solid

123

retention times of 13 h and 15 d, respectively. The hollow fiber membrane in use was

124

of hydrophilic polyvinylidene fluoride (PVDF), with a nominal pore size of 0.2 µm,

125

and was supplied by OriginWater (China).

126

AC C

EP

119

The overall process consisted of four parallel identical series, with the process flow 6

ACCEPTED MANUSCRIPT of each schematically shown in Fig. 1. In situ DO and ammonia-N probes (E+H,

128

Switzerland) were placed at the ends of aerobic tanks. Concentrations of DO and

129

ammonia-N were monitored via selective electrode analysis. An electromagnetic flow

130

meter (E+H, Switzerland) was placed at the influent pump to monitor the flow rate.

RI PT

127

Three high-speed centrifugal blowers (GL-TURBO, China) were installed for

132

aeration in aerobic tanks. Each blower could provide a maximum aeration flow rate of

133

4800 m3/h at a pressure of 68.6 kPa. The nominal power of each blower was 132 kW.

134

Output aeration flow rate could be adjusted by changing the opening angle of the

135

guide vane, achieving an aeration range of 60–100% of maximum flow rate. The

136

blowers were operated with one in service and two on standby at high temperatures,

137

and with two in service and one on standby at low temperatures, as determined by the

138

operators of the WWTP.

140

M AN U

TE D

139

SC

131

2.2 Software for modeling and simulation

The Activated Sludge Model 2d (ASM2d) was used to simulate the municipal

142

wastewater treatment process due to its suitable description of carbon oxidation,

143

nitrification, denitrification, biological and chemical phosphorus removal, and

144

denitrification by PAOs (Henze et al., 1999). Biowin software (Envirosim Corp.,

145

Canada), based on ASM2d, was used as the simulation platform in this study. The

146

membrane unit was described as a biological tank with additional parameters of

147

membrane rejection. The rejection rate of membrane for particles and colloids was set

148

as default values (100% and 95%, respectively) in this study. The software consisted

AC C

EP

141

7

ACCEPTED MANUSCRIPT 149

of simulation and control modules. Steady-state and dynamic-state simulation could

150

be achieved in the simulation module. Various control strategies (e.g., on/off, step,

151

and PI/PID) could be operated along with the process simulation.

153

2.3 Establishment of the full-scale MBR model

RI PT

152

The biological treatment process consisted of four parallel series, each of which

155

included an anoxic tank (A2), an anaerobic tank (A1), an interchangeable A/O tank (X),

156

an aerobic tank (O), and membrane tanks (M). The treatment capacity of each series

157

was 12500 m3/d (Fig. S1). Each tank was modeled according to its length, width, and

158

height. Sludge recirculation ratios were set according to actual operational conditions

159

(500% from M to O, 100% from O to A2, and 200% from O to X). A2 and A1 were set

160

up with no aeration (DO = 0 mg/L). X and O were set up with constant-flow aeration.

161

The total aeration flow rate was set at 2400 m3/h (in winter) and 1200 m3/h (in

162

summer), according to the actual operation of blowers. The aeration distribution ratio

163

of X to O was 0.15:0.85.

165 166

M AN U

TE D

EP

AC C

164

SC

154

2.4 Parameter calibration ASM2d involved 21 biological and chemical reaction processes, with 45 kinetic

167

parameters and 22 stoichiometric parameters (Henze et al., 1999). Sensitivity analysis

168

was conducted under steady-state simulation conditions to identify parameters that

169

significantly affected simulation results (Sui et al., 2014). These parameters were then

170

manually adjusted until the simulation result was acceptably close to experimental 8

ACCEPTED MANUSCRIPT 171

results. Experiments were conducted as follows. Continuous sampling and monitoring were conducted for one week. Two automatic

173

samplers were installed at the influent of the anoxic tank and at the effluent of the

174

membrane tank, respectively. The sampling intervals of influent and effluent were 1 h

175

and 2 h, respectively. All samples were characterized in terms of their chemical

176

oxygen demand (COD), biological oxygen demand (BOD5), ammonia-N, total

177

nitrogen (TN), phosphate, total phosphorus (TP), and suspended solids (SS). The

178

soluble fraction of influent was separated via 0.45-µm membrane filtration (HAWP,

179

Millipore, USA), and was characterized in terms of COD, TN, and TP. Mixed liquor

180

from every tank was sampled daily to measure MLSS and volatile MLSS (MLVSS)

181

concentrations. The soluble fraction of mixed liquor was separated as filtrate of

182

0.45-µm membrane filtration, and was characterized in terms of COD, ammonia-N,

183

TN, phosphate, and TP. All measurements were conducted according to standard

184

methods (S.E.P.A., 2002). Temperature, pH, and DO were measured daily using

185

portable devices (Thermo Orion, USA). In addition, the flow rate of influent was

186

recorded hourly by the central control system.

SC

M AN U

TE D

EP

AC C

187

RI PT

172

188

2.5 Aeration control strategy

189

2.5.1. Algorithm for the strategy

190

A two-step cascade control system was designed for aeration control in aerobic

191

tanks. The strategy commenced with obtaining in situ ammonia-N concentrations in

192

the aerobic tank and calculating the deviation of the ammonia-N setpoint from actual 9

ACCEPTED MANUSCRIPT values. The ammonia-N deviation was then used to calculate the DO deviation, and

194

thence the aeration demand was corrected. The whole calculation procedure from

195

ammonia-N to DO and then to aeration was performed in the PI framework, which

196

has been widely applied in industrial control systems (Piotrowski et al., 2012). PI

197

feedback control is to eliminate the error of target variable from a desired setpoint, by

198

altering the control variable. The calculation procedure from DO to aeration was

199

linked to that from ammonia-N to DO in a cascaded structure. Fig. 2 shows a

200

structural diagram of the control system, with the detailed procedure described below:

201

(i) Manually assign a setpoint of ammonia-N concentration (denoted as

203

SC

NH_setpoint) based on effluent requirements and temperature conditions; (ii) Measure real-time ammonia-N concentration (NH_RT), and calculate the ammonia-N error (e_NH):

205

e_NH = NH_setpoint − NH_RT ;

(iii) Run the first-step PI to obtain the setpoint of DO (DO_setpoint):

DO_setpoint = Bias1 + Kp1 ⋅ e_NH + K p1 ⋅

EP

207

TE D

204

206

M AN U

202

RI PT

193

1

τ1

∫

ti

ti −1

e_NH dt

where Bias1, Kp1, and τ1 are controller parameters, and ti−ti-1 is the control interval (30

209

min). The three terms on the right-hand side of the equation represent baseline,

210

real-time error, and historically accumulated error, respectively;

211

AC C

208

(iv) Measure the real-time DO concentration (DO_RT), and calculate the DO error

212

(e_DO):

213

e_DO = DO_setpoint − DO_RT

214

(v) Run the second-step PI to obtain the aeration flow rate (Q_air0): 10

ACCEPTED MANUSCRIPT 215

216

Q_air0 = Bias2 + Kp2 ⋅ e_DO + Kp2 ⋅

1

ti

τ 2 ∫t

e_DO dt

i −1

where Bias2, Kp2, and τ2 are controller parameters. (vi) Make corrections to variations in wastewater load (actual wastewater influent

218

flow rate, Q_influent, compared to the designed value, Q_design), to calculate the

219

aeration offset (Q_air_offset):

e_influent = Q_design − Q_influent  K ⋅e_influent e_influent ⋅ e_NH > 0 Q_air_offset =  inf 0 e_influent ⋅ e_NH ≤ 0 

SC

220

RI PT

217

where Kinf is the controller parameter (−2000 (m3-air/h)/(m3-influent/d)). The aeration

222

offset is positive in the case of excess ammonia-N and influent flow rate, and negative

223

in the case of under-setpoint ammonia-N and insufficient influent flow rate.

224

M AN U

221

(vii) Calculate the corrected aeration flow rate (Q_air): Q_air = Q_air0 + Q_air_offset

226

Q_air was transformed into suitable digital signals for the main control panel (MCP)

227

of the blowers, which controlled the angle of the guide vane in this study. The control

228

strategy was tested in Biowin software to optimize the parameters, and then in

229

engineering application to finely regulate parameters and evaluate the practical energy

230

saving performance.

232

EP

AC C

231

TE D

225

2.5.2. Optimization of algorithmic parameters

233

The already calibrated full-scale MBR model (cf. Sections 2.3 & 2.4) was

234

employed as a referee platform for the optimization of algorithmic parameters (i.e.,

235

Bias1, Kp1, τ1, Bias2, Kp2, and τ2) for the aeration control strategy. Optimization was 11

ACCEPTED MANUSCRIPT conducted by incorporating the control-strategy algorithm into the full-scale MBR

237

model, with parameters refined iteratively via tentative calculations. Theoretically, an

238

ideal set of parameters in this platform should result in perfectly constant ammonia-N

239

concentrations in the aerobic tank (i.e., e_NH = 0 over time lapse).

RI PT

236

Operationally, for the optimization, the influent conditions used in the simulation

241

were randomly generated, compared to the normal range of these parameters in the

242

WWTP. The algorithmic parameters were qualified as optimized when calculated

243

ammonia-N concentrations in the aerobic tank became acceptably stable.

247

M AN U

SC

240

248

2.5.3. Engineering application of the aeration control strategy

On the basis of optimized algorithmic parameters, energy saving potential was

245

evaluated by comparing aeration flow rates obtained from the simulation, with and

246

without the aeration control strategy.

TE D

244

The aeration control strategy was programed in C# language on Windows XP SP3

250

and .Net framework 4.0. The program was installed into the central control system of

251

the full-scale MBR. The program read the database to obtain essential data, calculated

252

the output, and transported the control signal to the blower control system. All the

253

read calculations and controls were available in real-time (within seconds).

AC C

254

EP

249

255

3 Results and discussion

256

3.1 Long-term monitoring of energy consumption of the full-scale MBR

257

The full-scale MBR has been in commission since January 2012. Stable pollutant 12

ACCEPTED MANUSCRIPT removal performance was achieved through the modified A2/O-MBR process (Table

259

S1). MBR effluent is able to meet the requirements of the National Standard I-A of

260

China, which specifies that concentrations of COD, BOD5, SS, TP, TN, and

261

ammonia-N should be below 50 mg/L, 10 mg/L, 20 mg/L, 0.5 mg/L, 15 mg/L, and 5

262

mg/L, respectively. Membrane filtration ensured high removal efficiency of COD and

263

SS. High biomass concentrations and prolonged SRT due to membrane rejection

264

contributed to good N removal performance.

SC

M AN U

265

RI PT

258

Energy consumption of the full-scale MBR during long-term operation was

267

monitored monthly (Fig. 3a). The MBR was operated at half capacity (25000 m3/d) in

268

2012 due to incomplete construction of the wastewater pipe network but has been

269

operated at full capacity (50000 m3/d) since 2013. Total monthly energy consumption

270

varied between 295.7–355.5 MWh in 2012, and between 549.3–648.8 MWh from

271

2013. Specific energy consumption has ranged between 0.35–0.65 kWh/m3-effluent

272

during the past three years. The average is 0.47 kWh/m3-effluent, a little lower than

273

the typical average specific energy consumption value of MBR in China (0.50

274

kWh/m3-effluent) obtained in a study by Xiao et al. (2014). Specific energy

275

consumption was a little higher than in Japan (0.39 kWh/m3-effluent) (Itokawa et al.,

276

2014) and Singapore (0.39 kWh/m3-effluent, under optimized conditions) (Tao et al.,

277

2010), and significantly lower than in Europe (0.8–2.4 kWh/m3-effluent) (Barillon et

278

al., 2013; Krzeminski et al., 2012) as reported. Specific energy consumption varied

279

seasonally, mostly maintained at a lower level from April to September and at a

AC C

EP

TE D

266

13

ACCEPTED MANUSCRIPT 280

higher level from November to January. The constitution of the energy consumption since 2012 was listed in Fig. 3b. Nearly

282

60% of the total power was consumed in aerobic tanks and membrane tanks. Aeration

283

for biological degradation and membrane scouring was the process of highest energy

284

consumption. The ratio of energy for biological degradation to membrane scouring

285

ranged within 0.85:1 in the full-scale MBR, suggesting that the aeration energy

286

consumption in aerobic tanks was comparable with that in membrane tanks. Hence,

287

reducing the energy consumption in aerobic would be of significance towards

288

lowering the overall energy consumption of the full-scale MBR.

M AN U

SC

RI PT

281

289 290

3.2 Model calibration

The full-scale MBR model was established on Biowin software and was calibrated

292

to the actual WWTP. The input of the model included influent characteristics and

293

operational conditions (e.g. SRT, DO concentration, etc.). The output of the model

294

included effluent characteristics and aeration demand in each tank, which was

295

calculated via numerically solving a series of differential equations.

EP

AC C

296

TE D

291

The simulation of the wastewater treatment process involved 21 biochemical

297

processes with 45 kinetic and 22 stoichiometric parameters. The default values of

298

these parameters were provided by the software, which were mostly obtained from

299

conventional activated sludge (CAS) processes. Since the full-scale MBR was

300

operated with longer SRT and lower F/M ratio, the default parameters of CAS might

301

be inappropriate for simulating the MBR process (Fenu et al., 2010). However, 14

ACCEPTED MANUSCRIPT measurement of each parameter was laborious due to complexity of experiments and

303

large amount of total parameters. Hence, sensitivity analysis was conducted to

304

identify the key parameters before the calibration. Under steady-state simulation with

305

constant influent characteristics (Table S2), variation of simulation output against the

306

variation of parameters was plotted in Fig. S2, suggesting significant impact of

307

maximum growth rate and aerobic decay rate of ammonia oxidation bacteria (AOB)

308

and nitrite oxidation bacteria (NOB) on ammonia simulation output, and maximum

309

growth rate, aerobic decay rate and yield coefficient of heterotrophic bacteria (HB) on

310

BOD, COD and TN simulation output (Machado et al., 2014).

M AN U

SC

RI PT

302

The selected parameters were manually calibrated according to experiments and

312

references (Table 1). The parameters of influent fractions differed from the default

313

value due to the induction of industrial wastewater, resulting in higher proportion of

314

soluble undegradable COD and TKN. The phosphate concentration in influent was

315

extremely low, which was possibly contributed by the aluminiferous wastewater from

316

electronic and mechanical manufacturing (Table S3). As to process parameters,

317

prolonged SRT and low F/M ratio, compared with CAS process, issued in decrease of

318

maximum growth rate and increase of decay rate of AOB, NOB and HB. The yield

319

coefficient of HB was also lower than the default value (Ruiz et al., 2013; Sui et al.,

320

2014). As to oxygen transfer process, the alpha factor was set to 0.75 according to an

321

oxygen-transfer experiment.

AC C

EP

TE D

311

322

The model simulation was conducted with calibrated parameters under

323

dynamic-state simulation condition. The input of the model was real influent 15

ACCEPTED MANUSCRIPT characteristics of 1-h interval obtained from a 4-d monitoring. The simulation results

325

was compared with experimental monitoring to evaluate the fitness of the model to

326

the full-scale MBR (Fig. 4). Concentration variations of COD, ammonia-N and TN

327

were of particular concern. The simulation result of COD acceptably matched the

328

experimental monitoring on the variation trend. The average deviation was 14.2 mg/L.

329

The fluctuation of simulation result was less drastic than that of experimental

330

monitoring. Since real influent contained irregular industrial wastewater, the actual

331

parameters of influent fractions might change. Nevertheless, parameters of influent

332

fractions could only set as constant value in the current model, leading to the

333

inaccurate simulation of COD variation. As to ammonia-N and TN removal, the

334

model showed good fitness. The average deviation in ammonia-N and TN removal

335

simulation was 0.36 mg/L (12%) and 0.44 mg/L (3%), respectively. TP removal was

336

not discussed in this study since TP and soluble TP concentrations were extremely

337

low in the influent (Table S3). Overall, the comparison of simulation results with

338

experimental monitoring indicated the validity of the model on wastewater treatment

339

process, especially ammonia-N removal process, which could provide a reliable base

340

for analyzing the energy saving potential of aeration control strategies.

342

SC

M AN U

TE D

EP

AC C

341

RI PT

324

3.3 Evaluation of energy saving potential of the aeration control strategy

343

The calibrated model (established in Section 3.2) could be used to simulate the

344

wastewater treatment process under certain influent conditions given constant

345

aeration flow rate. However, a drawback of the model is that the model cannot run 16

ACCEPTED MANUSCRIPT reversely to derive the real-time aeration demand from the influent conditions and

347

effluent requirements. As a result, aeration control and energy saving cannot be

348

achieved directly via the current model. Therefore, a straightforward relationship

349

(empirical and process-specific, but highly applicable) between aeration demand and

350

influent/effluent conditions, was required to establish based on the current model. A

351

feedback algorithm was implemented to extract the relationship from the model. The

352

algorithm started from effluent characteristics, and should be of simple calculation,

353

towards a real-time control of aeration in aerobic tanks.

M AN U

SC

RI PT

346

The aeration control strategy used in this study was an in-situ ammonia-N-based

355

feedback control with influent-flow-rate-based feedforward control. The detail of the

356

aeration control strategy was listed in Section 2.5. In order to achieve the control

357

strategy along with the simulation of wastewater treatment process, the control

358

strategy was programmed in Biowin language. The energy saving potential could be

359

analyzed by dint of process simulation with aeration control strategy. Under dynamic

360

simulation condition, the variation of aeration flow rate and ammonia-N concentration

361

in aerobic tank could be calculated from the influent characteristics, which was

362

random generated according to actual ranges of influent flow rate and pollutant

363

concentrations (Fig. S3). Simulation was conducted under two representative

364

temperatures of winter and summer, 13 and 26 °C. Aeration control strategy was

365

compared with constant aeration in terms of aeration flow rate, ammonia-N

366

concentration in aerobic tank, and ammonia-N and TN concentrations in effluent, to

367

evaluate the performance of the control strategy (Fig. 5). Aeration flow rate under

AC C

EP

TE D

354

17

ACCEPTED MANUSCRIPT aeration control strategy varied dynamically according to the simulation, and was

369

significantly lower than constant aeration flow rate, which represented current

370

operation condition, at both temperatures. Aeration flow rate was reduced by 15% and

371

20% in winter and summer, respectively. The ammonia-N concentration in aerobic

372

tank increased slightly, while that in effluent stably reached the standard (5 mg/L) due

373

to ammonia degradation in membrane tanks. The TN concentration also reached the

374

standard (15 mg/L), indicating that the aeration control strategy would have negligible

375

impact on N removal performance of the MBR. In summer particularly, the aeration

376

control strategy achieved a more stable ammonia-N concentration in aerobic tank and

377

effluent. The fluctuation of ammonia-N concentration, in terms of coefficient of

378

variation (standard deviation divided by average), decreased from 29.2% to 10.4%,

379

while that of influent was 19.2%. This suggested that the more stable effluent

380

characteristics could be achieved via the aeration control. However, the improvement

381

of the stability was not obvious in winter. The coefficient of variation only decreased

382

from 29.8% to 21.6%, which was possibly attributed to the impact of low temperature

383

on microorganism activity and kinetic parameters.

SC

M AN U

TE D

EP

AC C

384

RI PT

368

According the simulation, aeration flow rate could reduce by 15% to 20% via the

385

aeration control strategy. The simulation results suggested that the aeration control

386

strategy possessed a promising energy saving potential for full-scale engineering

387

application.

388

In current DO-based aeration control systems, accurate control of effluent

389

characteristics (e.g. ammonia-N) could not be achieved since DO was just an 18

ACCEPTED MANUSCRIPT intermediate variable. While the ammonia-N-based aeration control directly used

391

in-situ water characteristics as originate, leading to more accurate control of water

392

quality and more stable response to influent fluctuation. The key of the control

393

aeration strategy was two PI equations from ammonia-N to DO, and from DO to

394

aeration flow rate. The parameters were set as different values (see Table S4 in

395

Supplementary Material) and then dynamic simulations were conducted. The

396

simulation result, which was of both stable ammonia-N concentration variation and

397

appropriate aeration flow rate variation, indicated the suitable combination of

398

parameters. The optimized control parameters were: Kp1 = −2 (mgDO/L)/(mgN/L), τ1

399

= 20 min, Bias1 = 1 mg/L, Kp2 = 350 (m3-air/h)/(mgDO/L), τ2 = 15 min, Bias2= 2000

400

(13 °C) or 800 (26 °C) m3-air/h. This control strategy was a promising improvement

401

on current DO-based control, which was also verified through the simulation

402

comparison between DO-based control and ammonia-N-based control (Fig. 5). The

403

ammonia-N-based control strategy performed better in summer according to

404

simulation results. In winter, the ammonia-N concentration was occasionally over 5

405

mg/L at high influent load (Fig. 5(b)). The concentration in effluent was all below 5

406

mg/L due to the degradation of ammonia-N in membrane tanks (blue solid line in Fig.

407

5(c)).

408

could be achieved by optimization of aeration distribution of X and O tanks, sludge

409

recirculation from membrane tanks, etc.

AC C

EP

TE D

M AN U

SC

RI PT

390

While the performance in winter is still of improvement potential, which

410 411

3.4 Engineering application 19

ACCEPTED MANUSCRIPT In the engineering application of the control strategy, ammonia-N and DO

413

concentrations were monitored by in-situ meters. Aeration flow rate was calculated

414

and then was transformed into blower control signal, which was the angle of guide

415

vane of blowers in this full-scale MBR. The control signal was transferred to the MCP

416

of blowers in order that automatic control of blowers was achieved. The

417

above-mentioned was attained via programming using C# language. The program

418

performed functions of parameter configuration and controlled process calculation

419

(Fig. S4). The parameters in use was optimized as discussed in Section 3.3.

M AN U

SC

RI PT

412

Short-term monitoring of the full-scale MBR was performed during the operation

421

of the aeration control strategy for over one week, focusing on the aeration output of

422

blowers, actual power of blowers, and variation of ammonia-N concentration (Fig. 6).

423

After the aeration control strategy was applied, significant reduction of aeration flow

424

rate and power of blowers occurred. Aeration flow rate was reduced by 20%, while

425

the actual power of blowers for aerobic tank was reduced by 14%. The ammonia-N

426

concentration in aerobic tank increased but was still lower than the standard (5 mg/L).

427

Considering ammonia degradation in membrane tanks, the effluent quality could

428

safely meet the discharge requirement. The ammonia-N load in influent was similar

429

between the two weeks (5.02 ± 0.26 gN/(m3·d) in the first week and 5.22 ± 0.16

430

gN/(m3·d) in the second week). Consequently, the less variated influent conditions

431

might lead to a plus benefit on stable ammonia-N concentration aerobic tank, but the

432

reduction of aeration flow rate and actual power of blowers was attributed to the

433

operation of the aeration control strategy.

AC C

EP

TE D

420

20

ACCEPTED MANUSCRIPT Overall, the control strategy could function well in the full-scale MBR, resulting in

435

a specific energy consumption reduction of 4% in the whole wastewater treatment

436

process. In an early study, Verrecht et al. (2010) achieved a 23% reduction in energy

437

consumption, from 4.03 to 3.10 kWh/m3-effluent, by fixing the aeration intensity to a

438

lower level in relatively small MBR (25 m3/d). In the present study where the energy

439

consumption had already been as low as 0.47 kWh/m3-effluent, a further 4%

440

reduction was enabled by employing the real-time dynamic aeration control strategy.

441

Furthermore, a much higher energy reduction rate is expected when the present

442

strategy is applied to the cases of variable-frequency blowers (rather than

443

fixed-frequency blowers employed in this study).

M AN U

SC

RI PT

434

As to long-term operation, the aeration control strategy would lead to an annual

445

energy saving of 142 MWh. The price of power for industrial use in China was 0.6‒

446

0.9 RMB/kWh. And the cost of two ammonia-N probes were 120,000‒240,000 RMB.

447

The annual energy saving of 142 MWh would result in a payback period of this

448

control strategy is 0.9‒2.8 years, which was shorter than typical service time of a

449

well-maintained ammonia-N probe (3‒4 years). This suggested a worthy investment

450

on the aeration control system in large-scale MBR application. Comparing with

451

DO/ORP-based control and MLSS-based control, the ammonia-N-based aeration

452

control strategy achieved better effluent stability (Fatone et al., 2008; Zha et al., 2006).

453

However, the practical energy saving performance was limited in this full-scale

454

application. With consideration of safe operation of the full-scale MBR, the

455

adjustment amplitude of aeration flow rate was smaller than that in simulation.

AC C

EP

TE D

444

21

ACCEPTED MANUSCRIPT Moreover, the blowers applied in this full-scale MBR were of fixed-frequency

457

configuration. The rotating speed of motor was as a constant value once the blower

458

was turned on. Energy saving was achieved via altering the opening angel of guide

459

vanes and then the pressure loss. The percentage of power reduction was significantly

460

lower than that of aeration flow rate. As to variable-frequency blowers, the rotating

461

speed of motor could be altered based on the requirement of aeration output. The

462

aeration flow rate was adjusted via altering the frequency of the alternating current

463

and then the rotating speed, resulting in greater power reduction than fixed-frequency

464

blowers under same aeration reduction condition (Monclús et al., 2015). Thus, the

465

energy saving would be greater for those WWTPs where variable-frequency blowers

466

were applied. In general, engineering application in this study indicated that the

467

ammonia-N-based aeration control strategy could function well in full-scale MBR,

468

and possessed a promising application potential. The control strategy was also tested

469

effective by simulation of another large-scale MBR with different influent conditions

470

and treatment process in China. Besides, it is noteworthy that the stability of this

471

control system during long-time operation is of great importance and will be

472

investigated thoroughly in following studies.

SC

M AN U

TE D

EP

AC C

473

RI PT

456

474

4 Conclusion

475

This study proposed an aeration control strategy for aerobic tanks of MBR based on a

476

two-step cascaded PI algorithm, and optimized the algorithmic parameters via

477

simulation on a calibrated full-scale MBR model. The strategy was applied in model 22

ACCEPTED MANUSCRIPT platform, and the simulation indicated an aeration reduction potential of 15 to 20%

479

under the condition of the present MBR. The feasibility of the strategy in the

480

full-scale MBR was verified via engineering application, indicating 20% reduction of

481

aeration flow rate, and 4% reduction of overall specific energy consumption in the

482

case of fixed-frequency blowers. Greater reduction of energy consumption is

483

promising for those MBR installed with variable-frequency blowers.

RI PT

478

SC

484

Acknowledgement

486

This work was supported by the Program for Outstanding PhD thesis of Beijing

487

(No.2131000306) and the Major Science and Technology Program for Water

488

Pollution Control and Treatment (No.2011ZX07301-002). The authors also gave

489

thanks to Mr. Mingda Qian and Mr. Wei Liu from the WWTP for the assistance for

490

providing basic information of the MBR and helping conducting the experiments. The

491

authors additionally acknowledged Dr. Xiaolin Li from OriginWater Corp. and Mr.

492

Xinglin Ge from Wuxi GL-TURBO Corp. for the assistance in programming and

493

debugging in the engineering application.

TE D

EP

AC C

494

M AN U

485

495

References

496

Barillon, B., Ruel, S.M., Langlais, C., Lazarova, V. 2013. Energy efficiency in

497

membrane bioreactors. Water Sci. Technol., 67(12), 2685-2691.

498

Brepols, C., Schafer, H., Engelhardt, N. 2010. Considerations on the design and

499

financial feasibility of full-scale membrane bioreactors for municipal 23

ACCEPTED MANUSCRIPT 500

applications. Water Sci. Technol., 61(10), 2461-2468. Fatone, F., Battistoni, P., Bolzonella, D., Pavan, P., Cecchia, F. 2008. Long-term

502

experience with an automatic process control for nitrogen removal in

503

membrane bioreactors. Desalination, 227(1–3), 72-84.

RI PT

501

Fenu, A., Guglielmi, G., Jimenez, J., Spèrandio, M., Saroj, D., Lesjean, B., Brepols,

505

C., Thoeye, C., Nopens, I. 2010. Activated sludge model (ASM) based

506

modelling of membrane bioreactor (MBR) processes: A critical review with

507

special regard to MBR specificities. Water Res., 44(15), 4272-4294.

M AN U

SC

504

508

Gabarron, S., Dalmau, M., Porro, J., Rodriguez-Roda, I., Comas, J. 2015.

509

Optimization of full-scale membrane bioreactors for wastewater treatment

510

through a model-based approach. Chem. Eng. J., 267, 34-42. Hauduc, H., Rieger, L., Oehmen, A., van Loosdrecht, M.C.M., Comeau, Y., Heduit, A.,

TE D

511

Vanrolleghem, P.A., Gillot, S. 2013. Critical review of activated sludge

513

modeling: State of process knowledge, modeling concepts, and limitations.

514

Biotechnol. Bioeng., 110(1), 24-46.

516 517 518

Henze, M., Gujer, W., Mino, T., Matsuo, T., Wentzel, M.C., Marais, G.V.R., Van

AC C

515

EP

512

Loosdrecht, M.C.M. 1999. Activated Sludge Model No.2d, ASM2d. Water Sci. Technol., 39(1), 165-182.

Huang, X., Xiao, K., Shen, Y.X. 2010. Recent advances in membrane bioreactor

519

technology for wastewater treatment in China. Front. Environ. Sci. Eng. China,

520

4(3), 245-271.

521

Ingildsen, P., Jeppsson, U., Olsson, G. 2002. Dissolved oxygen controller based on 24

ACCEPTED MANUSCRIPT 522

on-line measurements of ammonium combining feed-forward and feedback.

523

Water Sci. Technol., 45(4-5), 453-460. Itokawa, H., Tsuji, K., Yamashita, K., Hashimoto, T. 2014. Design and operating

525

experiences of full-scale municipal membrane bioreactors in Japan. Water Sci.

526

Technol., 69(5), 1088-1093.

528

Judd, S., Judd, C. 2011. The MBR Book: Principles and Applications of Membrane Bioreactor for Water and Wastewater Treatment. 2nd ed. Elsevier, Oxford.

SC

527

RI PT

524

Krzeminski, P., van der Graaf, J., van Lier, J.B. 2012. Specific energy consumption of

530

membrane bioreactor (MBR) for sewage treatment. Water Sci. Technol., 65(2),

531

380-392.

M AN U

529

Machado, V.C., Lafuente, J., Baeza, J.A. 2014. Activated sludge model 2d calibration

533

with full-scale WWTP data: comparing model parameter identifiability with

534

influent and operational uncertainty. Bioprocess. Biosyst. Eng., 37(7),

535

1271-1287.

TE D

532

Monclús, H., Dalmau, M., Gabarrón, S., Ferrero, G., Rodríguez-Roda, I., Comas, J.

537

2015. Full-scale validation of an air scour control system for energy savings in

AC C

538

EP

536

membrane bioreactors. Water Res., 79(0), 1-9.

539

Piotrowski, R., Zawadzki, A., Ieee. 2012. Multiregional PI control strategy for

540

dissolved oxygen and aeration system control at biological wastewater

541

treatment plant. 2012 IEEE 17th Conference on Emerging Technologies &

542

Factory Automation (Etfa), 8.

543

Pittoors, E., Guo, Y.P., Van Hulle, S.W.H. 2014. Modeling Dissolved Oxygen 25

ACCEPTED MANUSCRIPT 544

Concentration for Optimizing Aeration Systems and Reducing Oxygen

545

Consumption in Activated Sludge Processes: A Review. Chem. Eng. Commun.,

546

201(8), 983-1002. Ruiz, L.M., Perez, J., Gomez, M.A. 2013. Influence of operational parameters over

RI PT

547 548

biomass growth and decay kinetic constants on membrane bioreactors. Desalin.

549

Water Treat., 51(25-27), 4847-4854.

SC

551

S.E.P.A. 2002. Standard Methods for the Examination of Water and Wastewater. 4th ed. China Environmental Science Press, Beijing.

M AN U

550

552

Sui, J., Li, J., Zhang, F. 2014. Optimization of Design and Operation of Sewage

553

Treatment Process by Process Simulation System. China Water & Wastewater,

554

30(11), 111-115.

Sun, J., Xiao, K., Mo, Y., Liang, P., Shen, Y., Zhu, N., Huang, X. 2014. Seasonal

556

characteristics of supernatant organics and its effect on membrane fouling in a

557

full-scale membrane bioreactor. J. Membr. Sci., 453, 168-174.

TE D

555

Sun, J., Xiao, K., Yan, X., Liang, P., Shen, Y.-x., Zhu, N., Huang, X. 2015. Membrane

559

bioreactor vs. oxidation ditch: full-scale long-term performance related with

AC C

560

EP

558

mixed liquor seasonal characteristics. Process Biochem., 50, 2224–2233.

561

Tao, G., Kekre, K., Oo, M.H., Viswanath, B., Yusof, A.M., Seah, H. 2010. Energy

562

Reduction and Optimisation in Membrane Bioreactor Systems. Water Pract.

563

Technol., 5(4), 7.

564

Verrecht, B., Maere, M., Benedetti, L., Nopens, I., Judd, S. 2010. Model-based energy

565

optimisation of a small-scale decentralised membrane bioreactor for urban 26

ACCEPTED MANUSCRIPT 566

reuse. Water Res., 44(14), 4047-4056. Wang, C., Zeng, Y.Z., Lou, J., Wu, P. 2007. Dynamic simulation of a WWTP operated

568

at low dissolved oxygen condition by integrating activated sludge model and a

569

floc model. Biochem. Eng. J., 33(3), 217-227.

RI PT

567

Xiao, K., Xu, Y., Liang, S., Lei, T., Sun, J.Y., Wen, X.H., Zhang, H.X., Chen, C.S.,

571

Huang, X. 2014. Engineering application of membrane bioreactor for

572

wastewater treatment in China: Current state and future prospect. Front. Env.

573

Sci. Eng., 8(6), 805-819.

M AN U

SC

570

Yan, X., Xiao, K., Liang, S., Lei, T., Liang, P., Xue, T., Yu, K., Guan, J., Huang, X.

575

2015. Hydraulic optimization of membrane bioreactor via baffle modification

576

using computational fluid dynamics. Bioresour. Technol., 175, 633-637.

577

Zha, F., Liu, W., Jordan, E., Kuzma, M., Jordan, E.J., Kuchlma, M., 2006. Controlling

578

operating parameters of membrane bioreactor system involves determining

579

control algorithm and controlling operating parameter(s) using determined

580

control algorithm, WO2007038843-A1 WOAU001472 06 Oct 2006.

EP

AC C

581

TE D

574

27

ACCEPTED MANUSCRIPT Figure captions

583

Fig. 1. Schematic process diagram of the full-scale MBR. The interchangeable A/O

584

tank can be operated under either anoxic or aerobic condition, depending on practical

585

circumstances.

586

Fig. 2. Structural diagram of the two-step proportion-integration cascade control

587

system for aeration control in the aerobic tank.

588

Fig. 3. Long-term monitoring of the energy consumption and its constitution in the

589

full-scale MBR.

590

Fig. 4. Experimental monitoring of influent and effluent concentrations of COD (a),

591

ammonia-N (b), and TN (c) compared with simulation results.

592

Fig. 5. Simulation of the aeration control strategy on Biowin software under random

593

influent condition. Performances of constant aeration, DO-based aeration control, and

594

ammonia-N-based control were simulated at summer and winter temperatures. The

595

aeration flow rate, ammonia-N concentration in aerobic tank, and effluent quality at a

596

winter temperature (13 °C) are shown in (a), (b) and (c); those at a summer

597

temperature (26 °C) are shown in (d), (e) and (f).

598

Fig. 6. Profile of aeration flow rate, actual power of blowers, ammonia-N

599

concentration in aerobic tank and effluent during the operation of the aeration control

600

strategy.

AC C

EP

TE D

M AN U

SC

RI PT

582

601

28

ACCEPTED MANUSCRIPT Table captions

603

Table 1. Calibrated parameters for the simulation in Biowin software.

AC C

EP

TE D

M AN U

SC

RI PT

602

29

ACCEPTED MANUSCRIPT Table 1. Calibrated parameters for the simulation in Biowin software. Unit

Default value

Calibrated value

gCOD/gTCOD

0.05

0.07

g-Ammoia-N/gTKN

0.66

0.495

Soluble undegradable CODa Ammoniaa Particle organic Na

0.50

0.394

0.5

0.022

0.02

0.08

0.9

0.72

d-1

0.17

0.23

d-1

0.70

0.50

d-1

0.17

0.23

d-1

3.20

2.80

d-1

0.62

0.70

d-1

0.23

0.25

Yield coefficient (HB)b

-

0.666

0.40

Alpha factorc

-

0.5

0.75

Phosphatea

g-Phosphate-P/gTP

Soluble undegradable TKNa

g-Ammonia-N/gTKN

Aerobic decay rate (AOB)b Maximum growth rate (NOB)b Aerobic decay rate (NOB)b

Aerobic decay rate (HB)b

AC C

EP

Anoxic decay rate (HB)b

TE D

Maximum growth rate (HB)b

d-1

M AN U

Maximum growth rate (AOB)b

SC

g-Ammonia-N/g-Organic-N

RI PT

Parameter

Abbreviations: AOB, ammonia oxidation bacteria; COD, chemical oxygen

demand; HB, heterotrophic bacteria; NOB, nitrate oxidation bacteria; TKN, total Kjeldahl nitrogen; TP, total phosphorus. a

b

Parameters were calibrated according to experimental tests of influent. Parameters were manually calibrated around values obtained from former

researches and local experiments.

ACCEPTED MANUSCRIPT c

Parameter was calibrated according to an oxygen-transfer experiment

AC C

EP

TE D

M AN U

SC

RI PT

(unpublished data).

ACCEPTED MANUSCRIPT (in situ ammonia-N and DO probes)

Primary effluent

NH DO Anoxic tank

Anaerobic Interchangeable A/O tank tank

Aerobic tank

Membrane tank

Sludge recirculation

RI PT

Effluent

Excess sludge

Fig. 1. Schematic process diagram of the full-scale MBR. The interchangeable A/O

SC

tank can be operated under either anoxic or aerobic condition, depending on practical

AC C

EP

TE D

M AN U

circumstances.

ACCEPTED MANUSCRIPT e_NH NH_RT

(1) In situ ammonia probe (2) In situ DO probe (3) Error calculator of ammonia (4) Ammonia-DO PI controller (5) Error calculator of DO (6) DO-aeration PI controller (7) Influent flow rate meter (8) Influent-aeration offset controller (9) Aerobic tank (10) Blower for aerobic tank

(5)

NH-DO PI

DO_setpoint

(6) e_DO

+

-

DO-Aeration PI Q_air0

DO_RT NH

DO

(1)

(2) Q_air (9)

(10) (7)

Q

RI PT

NH_setpoint +

(4)

Q_air_offset

Qinfluent Influent-aeration offset (8)

SC

(3)

M AN U

Fig. 2. Structural diagram of the two-step proportion-integration cascade control

AC C

EP

TE D

system for aeration control in the aerobic tank.

ACCEPTED MANUSCRIPT 1000

0.5

600

0.4 400

0.3 0.2

200

Half capacity 25000 m3/d

Grit chamber & Fine screen 6.8% Super-fine screen 1.5% Anoxic tank 3.5% Anaerobic tank 1.8% Simultaneous A/O tank 3.2%

Membrane tank 32.1%

Full capacity 50000 m3/d

0.0

Influent 14.5%

RI PT

800 0.6

Total energy consumption (MWh)

0.7

0.1

Effluent 10.6%

Specific energy consumption Total energy consumption

0

Aerobic tank 26.0%

SC

Jan Apr Jul Oct Jan Apr Jul Oct Jan Apr Jul Oct 2012 2013 2014 Date

Fig. 3. Long-term monitoring of the energy consumption and its constitution in the

EP

TE D

M AN U

full-scale MBR.

AC C

Specific energy consumption (kWh/m3-effluent)

0.8

ACCEPTED MANUSCRIPT

(b) Ammonia

20 10

0

24

20 10

AC C

0

EP

140 (c) TN 120 100 80 60 40

0

RI PT

48 Time (h)

30

0

Concentration (mg/L)

24

72

M AN U

Concentration (mg/L)

40

0

Effluent (Simulation)

96

SC

80 60 40 20 0

Effluent (Experiment)

48 Time (h)

72

96

72

96

TE D

Concentration (mg/L)

Influent (Experiment) 1200 1000 (a) COD 800 600 400 200

24

48 Time (h)

Fig. 4. Experimental monitoring of influent and effluent concentrations of COD (a), ammonia-N (b), and TN (c) compared with simulation results.

ACCEPTED MANUSCRIPT 1500

2000 1500

Average Average Constant aeration Ammonia-N-based control DO-based control

500

1200 900 600

0 48

72 96 Time (h)

120

144

168

0

4

(b) Ammonia concentration in aerobic tank (13 °C) Concentration (mg/L)

5 4 3

1 0

24

48

72 96 Time (h)

120

144

168

(e) Ammonia concentration in aerobic tank (26 °C)

3

2

1

M AN U

2

0

0

24

48

72 96 Time (h)

120

144

0 48

72 96 Time (h)

120

144

EP

24

24

48

72 96 Time (h)

120

144

168

120

144

168

(f) Effluent quality (26 °C) Solid line: Ammonia; Dash line: TN

10

168

5

0 0

24

48

72 96 Time (h)

Fig. 5. Simulation of the aeration control strategy on Biowin software under random

AC C

0

TE D

15

5

0

15

(c) Effluent quality (13 °C) Solid line: Ammonia; Dash line: TN

10

168

Concentration (mg/L)

Concentration (mg/L)

24

SC

0

20

Average Average

300

0

6

(d) Sum of aeration flow rate (26 °C)

RI PT

1000

Aeration flow rate (m3/h)

(a) Sum of aeration flow rate (13 °C)

2500

Concentration (mg/L)

Aeration flow rate (m3/h)

3000

influent condition. Performances of constant aeration, DO-based aeration control, and ammonia-N-based control were simulated at summer and winter temperatures. The aeration flow rate, ammonia-N concentration in aerobic tank, and effluent quality at a winter temperature (13 °C) are shown in (a), (b) and (c); those at a summer temperature (26 °C) are shown in (d), (e) and (f).

ACCEPTED MANUSCRIPT

200

3600

150

2400

100

Aeration flow rate in biological tanks Actual power

0 4

0

(b)

Ammonia N concentration in aerobic tank (mg/L)

SC

3 2

0

2

4

M AN U

1 0

50

RI PT

1200 (a)

Actual power (kW)

Ammonia-N-based aeration control

4800

Ammonia N concentration in aerobic tank (mg/L)

Aeration flow rate in biological tanks (m3/h)

Constant aeration flow rate

6

8 Time (d)

10

12

14

16

Fig. 6. Profile of aeration flow rate, actual power of blowers, ammonia-N

AC C

EP

strategy.

TE D

concentration in aerobic tank and effluent during the operation of the aeration control

ACCEPTED MANUSCRIPT
A model with calibrated parameters was established to simulate a full-scale MBR.
A two-step cascaded PI algorithm was proposed for aeration control in aerobic tank.

RI PT


Aeration could be reduced by 15–20% according to simulation of the algorithm.
Aeration and energy consumption was reduced by 20% and 4% in the full-scale

AC C

EP

TE D

M AN U

SC

MBR.