Linköpings universitet SE–581 83 Linköping
2018 | LITH-IFM-A-EX--18/3490--SE
Ozonation of pharmaceutical
residues in a wastewater
treatment plant
–
Modeling the ozone demand based on a multivariate
analysis of influential parameters
Ozonering av läkemedelsrester på ett avloppsreningsverk
-Modellering baserat på en multivariat analys av parametrar
som påverkar ozonbehovet
Emilia Johansson
Erica Engberg
Supervisor : Robert Gustavsson Examiner : Carl-Fredrik Mandenius
Department of Physics, Chemistry and Biology
Linköping University
URL för elektronisk version
ISBN
ISRN: LITH-IFM-A-EX--18/3490--SE
_________________________________________________________________
Serietitel och serienummer ISSN
Title of series, numbering ______________________________
Språk Language Svenska/Swedish Engelska/English ________________ Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport _____________ Titel
Ozonation of pharmaceutical residues in a wastewater
- Modeling the ozone demand based on multivariate analysis of influential parameters
Författare
Emilia Johansson Erica Engberg
Nyckelord Sammanfattning
Most pharmaceutical residues in wastewater treatment plants (WWTPs) end up in the hydrosphere where they cause negative effects on the aquatic life and might disrupt ecosystems. By implementing an ozonation step (treatment with ozone) in the wastewater treatment process, these pharmaceutical residues can be reduced. The purpose of this project was to verify that the ozonation process works in full-scale, thereby verifying a pilot study conducted in 2014 at Tekniska Verken i Linköping AB (TVAB). Additionally, the purpose was to investigate which parameters influence the ozone demand in order to formulate a model for the ozone demand. The initial phases during this thesis were a pre-study and a literature study. This was followed by the multivariate analysis and model construction based on different data from the pilot study. Measurements were performed on the wastewater in the full-scale facility in order to verify the results from the pilot study. Moreover, measurements were performed to find new ozone consuming parameters. The reduction of pharmaceutical residues was similar to the pilot study, although slightly lower. Several parameters and factors that were different between pilot study and new measurements affected the reduction of pharmaceutical residues. For example, DOC and nitrate concentrations have increased since the pilot study in 2014. Also, factors such as the growth in population in Linköping and the differences in design between the pilot plant and the full-scale facility have influenced the reduction of pharmaceutical residues. A control strategy based on a linear relationship between ozone sensitive Ultra Violet Absorption (UVA) left and remaining pharmaceutical residues after ozonation could potentially be used. Moreover, three models were constructed and the Multivariate Analysis 1 (MVA1)-model was deemed as the best, this model includes ozone residual, nitrite, turbidity, simulated Chemical Oxygen Demand (COD(sim)) and ozone dose. The variations in the dose compared to the input parameters for the validation data show that the model predict the ozone dose well. However, in future other interesting parameters can be included in the model to further improve the accuracy in the ozone dose predicted by the model.
Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – under 25 år från publiceringsdatum under förutsättning att inga extraordinära omständigheter uppstår. Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och admin-istrativ art. Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sam-manhang som är kränkande för upphovsmannens litterära eller konstnärliga anseende eller egenart. För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.
Copyright
The publishers will keep this document online on the Internet – or its possible replacement – for a period of 25 years starting from the date of publication barring exceptional circum-stances. The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the con-sent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility. According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement. For additional information about the Linköping Uni-versity Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.
c
Emilia Johansson Erica Engberg
Most pharmaceutical residues in wastewater treatment plants (WWTPs) end up in the hy-drosphere where they cause negative effects on the aquatic life and might disrupt ecosys-tems. By implementing an ozonation step (treatment with ozone) in the wastewater treat-ment process, these pharmaceutical residues can be reduced. The purpose of this project was to verify that the ozonation process works in full-scale, thereby verifying a pilot study conducted in 2014 at Tekniska Verken i Linköping AB (TVAB). Additionally, the purpose was to investigate which parameters influence the ozone demand in order to formulate a model for the ozone demand. The initial phases during this thesis were a pre-study and a literature study. This was followed by the multivariate analysis and model construc-tion based on different data from the pilot study. Measurements were performed on the wastewater in the full-scale facility in order to verify the results from the pilot study. More-over, measurements were performed to find new ozone consuming parameters. The reduc-tion of pharmaceutical residues was similar to the pilot study, although slightly lower. Sev-eral parameters and factors that were different between pilot study and new measurements affected the reduction of pharmaceutical residues. For example, DOC and nitrate concen-trations have increased since the pilot study in 2014. Also, factors such as the growth in population in Linköping and the differences in design between the pilot plant and the full-scale facility have influenced the reduction of pharmaceutical residues. A control strategy based on a linear relationship between ozone sensitive Ultra Violet Absorption (UVA) left and remaining pharmaceutical residues after ozonation could potentially be used. More-over, three models were constructed and the Multivariate Analysis 1 (MVA1)-model was deemed as the best, this model includes ozone residual, nitrite, turbidity, simulated Chem-ical Oxygen Demand (COD(sim)) and ozone dose. The variations in the dose compared to the input parameters for the validation data show that the model predict the ozone dose well. However, in future other interesting parameters can be included in the model to further improve the accuracy in the ozone dose predicted by the model.
We would like to thank TVAB and the department of Vatten och Avlopp for providing the opportunity of this thesis on a new and exiting matter. Secondly, we express our deepest gratitude to Robert Sehlén, our supervisor at TVAB, for valuable input, advice, guidance and never fading interest even when things did not go as planned. Thirdly, we thank our LiU supervisor Robert Gustavsson for his willingness to help and quick response no matter what the question might have been. Fourthly, we thank our examiner from LiU, Carl-Fredrik Mandenius for allocating some of his precious time to this thesis and for his input. Fiftly, we acknowledge the laboratory at TVAB, the laboratory at Aarhus university and SYNLAB for running measurements for us. Also, we thank our opponents for their suggestions on how to improve this report. Lastly, we express our gratitude to Maja Ekblad, Ulf Miehe, Michael Stapf and Alexander Sanner for valuable discussions, input and support during this thesis.
COD- Chemical Oxygen Demand
COD(sim)- Simulated Chemical Oxygen Demand CWPharma- Clear Waters from Pharmaceuticals DOC- Dissolved Organic Carbon
Fe2+- Bivalent Iron
FNU- Formazin Nephelometric Unit MBBR- Moving Bed Biofilm Reactor mS/m- mikroSiemens/meter
MTE- Mass Transfer Efficiency MVA- Multivariate Analysis N-model- Nitrogen Model NO2-N- Nitrite Nitrogen
PCA- Principal Component Analysis PLS- Partial Least Squares
TVAB- Tekniska Verken AB UVA- Ultra-Violet Absorption WWTP- Wastewater Treatment Plant MS-sim- MATLAB Simulink- simulation
Abstract iv Acknowledgments v Abbreviations vi Contents vii List of Figures ix List of Tables x 1 Introduction 1
1.1 Purpose of the study . . . 1
1.2 Objectives of the work . . . 2
1.3 Expected impact of the study . . . 2
1.4 Delimitations . . . 3
2 Theory and Methodology 4 2.1 Scientific background . . . 4
2.1.1 Ozone . . . 4
Generating ozone . . . 5
Reaction mechanism of ozone . . . 5
Parameters that influence the ozonation of pharmaceutical residues . . . 6
2.1.2 Pharmaceuticals . . . 8
Pharmaceuticals to be monitored . . . 8
Effects of pharmaceuticals . . . 9
2.1.3 Wastewater treatment . . . 11
Nykvarnsverket . . . 11
Pharmaceutical residue treatment . . . 12
2.1.4 Pilot study . . . 14 Background . . . 14 Method . . . 14 Results . . . 15 Possible Improvements . . . 16 2.2 Methodology . . . 17 2.2.1 Pre-study . . . 17 2.2.2 Literature study . . . 17 2.2.3 Main activities . . . 18
Multivariate analysis and regression . . . 18
Measurements on full-scale facility . . . 18
2.2.4 Validation . . . 19
2.3 Models . . . 20
2.3.1 PCA . . . 20
2.3.2 PLS . . . 20
3 Method and Materials 21 3.1 Method and materials . . . 21
3.1.1 Data analysis . . . 21
3.1.2 Multivariate analysis . . . 23
PCA . . . 23
PLS . . . 23
Regression - fitted line plot . . . 23
3.1.3 Model validation and testing . . . 23
Model validation . . . 23
Model testing . . . 24
3.1.4 Laboratory work . . . 24
Sampling days . . . 24
Ozonation . . . 25
3.1.5 Comparison between the pilot study and new measurements . . . 26
4 Results and Discussion 27 4.1 Verification of the pilot study . . . 27
4.1.1 Parameters . . . 27
4.1.2 Gas flow . . . 32
4.1.3 Pharmaceutical residues . . . 33
4.2 Model construction . . . 38
4.2.1 Multivariate analysis and regression model . . . 38
MVA-models . . . 38
Nitrite model . . . 39
4.2.2 Model selection and validation . . . 41
4.2.3 Model testing . . . 44
The procedure . . . 44
Ozone Dose(sim) validation . . . 45
Verification . . . 47
4.3 Additional aspects to consider . . . 48
4.3.1 New parameters . . . 48 4.3.2 Model improvements . . . 48 5 Conclusion 50 5.1 Conclusion . . . 50 5.2 Future Work . . . 52 Bibliography 53 A Appendix A 58 A.1 MATLAB script . . . 58
B Appendix B 59 B.1 Purpose, objectives and boundary conditions . . . 59
2.1 The structure of A) Diclofenac, B) Metoprolol and C) Oxazepam . . . 9 2.2 The wastewater treatment process at Nykvarnsverket. . . 12 2.3 The ozonation process with ozone generation from liquid oxygen and on-line
mea-surement points (On 1, On 2 and On 3). . . 13 2.4 A timeline illustrating how the work was carried out. . . 17 4.1 Daily variations of COD(sim) extracted from Linköpingsmodellen and measured
variations in DOC, UVA, nitrite and flow during the pilot study. . . 28 4.2 The ozone sensitive UVA left against A) The ozone dose, B) The ozone dose
di-vided by DOC concentration C) The ozone dose didi-vided by nitrite concentrations. 31 4.3 The ozone residual against the ozone dose for the dose-, control-, and repeat trials
with exponential trend lines. . . 32 4.4 The reduction of A) Diclofenac, B) Metoprolol and C) Oxazepam during the three
trials; dose-, control- and repeat trials of the pilot study and the reduction from measurements made on the full-scale facility. Sadly, oxazepam could not be mea-sured with the method used. . . 35 4.5 The average for diclofenac, metoprolol and oxazepam remaining after ozonation
against the ozone sensitive UVA left for the different trials and the new measure-ments as well as a linear trendline for the dose trial. The black arrow mark the outlier. . . 36 4.6 A) The measured ozone dose and the nitrite concentration over the day that the
N-model was based on. B) The measured ozone dose compared to the dose predicted by the N-model during the days when the regulatory strategy was deployed. . . . 40 4.7 The measured ozone residual compared to the residual predicted by the
MVA1-model and MVA2-MVA1-model for hourly averages of 15 days during the summer, where the ozone dose was kept at 9.75˘0.3 mg/L. . . 42 4.8 The measured ozone residual compared to MVA1-model residual for the data used
to construct the MVA1-model. . . 42 4.9 An overview of the Ozone Dose(sim) model . . . 45 4.10 The ozone dose from the Ozone Dose(sim)-model and measured A) Nitrite, B)
Turbidity and C) COD(sim) variations for two days; 2014-07-19 and 2014-07-27, during the 15 day validation period used. . . 46 4.11 The ozone dose predicted by the Ozone Dose(sim) (in mg/L on primary y-axis)
the turbidity (in FNU on primary y-axis) as well as the nitrite concentration (in mg/L on secondary y-axis) for three sets of input parameters (x-axis) from the new measurements. . . 47
3.1 Available data from measurements made and also measurements and on-line data from the pilot study. . . 21 3.2 Parameters that were measured before and after the ozonation step. . . 25 4.1 Daily averages of on-line data for the pilot study and the new measurements on
the ozone doses of 5, 6.5 and 8 mg/L. The change was calculated according to equation 3.1. . . 28 4.2 Parameters that influence the ozonation measured during the pilot study as well
as new measurements made on the ozone doses of 5, 6.5 and 8 mg/L. The change was calculated according to equation 3.1. . . 29 4.3 The nitrite concentration before and after ozonation at the tested ozone doses. The
reduction was calculated according to equation 3.2. . . 30 4.4 UVA results from measurements before and after the ozonation on the full-scale
ozonation.The ozone sensitive UVA left was calculated according to equation 3.3. . 30 4.5 Reduction of diclofenac and metoprolol during the new measurements on ozone
doses of 5, 6.5 and 8 mg/L. Sadly, oxazepam could not be measured with the method used. The reduction was calculated accoring to equation 3.2. . . 34 4.6 Analysis of Variance and Model-, and Validation selection for MVA1 and MVA2. . 38 4.7 The coefficients received from MVA1 and MVA2 for the analyzed parameters. . . . 39 4.8 Nitrite and nitrate concentrations before ozonation for the different doses as well
as calculated quotients and averages. . . 43 4.9 Parameters that are expected to influence the ozonation measured during the pilot
The trend for the annual sale of pharmaceuticals in Sweden is an increase in sales revenue as well as in defined daily doses[1]. This trend is estimated to continue as the population and the longevity increase [2] [3]. Pharmaceuticals improve the quality of life for the individual but there are environmental consequences to take into consideration. Most pharmaceutical residues are excreted in urine and excrements before reaching the wastewater treatment plant (WWTP). [4] Nowadays, no purification step to remove pharmaceutical residues is present in the WWTPs in Sweden [4, 5]. Therefore, the pharmaceutical residues reach the hydrosphere, where they have been shown to have negative effects on the aquatic life, for example caus-ing sterility in fish and thereby disruptcaus-ing ecosystems [4, 6]. Since many aquatic organisms, including fish, have similar target molecules such as receptors and enzymes, there is an im-minent danger that the long-term effects of the pharmaceutical residues may impact humans as well. Moreover, the effects on humans of the pharmaceutical residues in the tap water, as well as in seafood, are currently unknown, but might arise in the future. [5] By implementing an ozonation step for reduction of the pharmaceutical residues at WWTPs there will hope-fully never be an impact on humans and the observed negative effects on the aquatic life and environment will be minimized [4, 7].
1.1
Purpose of the study
Pharmaceutical residues in wastewater have been shown to have effects on the aquatic life and together with a political interest from Linköping municipality, Tekniska Verken i Linköping AB (TVAB) investigated the possibilities to reduce the loads of pharmaceutical residues. In a pilot study in 2014, the method ozonation (treatment with ozone) was tested. With the positive results, TVAB decided to implement a full-scale facility at Nykvarnsverket, Linköping. However, as ozonation of wastewater is a fairly unfamiliar process and wastew-aters differ among the WWTPs there is no standard operating procedure. TVAB has a goal to reduce 90 % of the pharmaceutical residues, in order to eliminate the negative effects of phar-maceutical residues on the environment, in the wastewater using ozonation and to achieve this the right ozone dose must be used. [8] Underdosing may result in insufficient reduction of pharmaceutical residues while overdosing of ozone may result in more oxidation products and unnecessary expenses. [9]
The purpose of this project was therefore to verify that the ozonation process works in full-scale, thereby verifying the pilot study and to investigate which parameters influence the ozone consumption in order to formulate a model for the ozone demand.
1.2
Objectives of the work
The main objectives were to verify the pilot study and to find the optimal ozone dose to reach a reduction of pharmaceutical residues of 90 % at different conditions in Nykvarnsverket. This was achieved through measurements, by investigating which parameters influence the ozone demand and the correlation between these parameters. Also, by developing and eval-uating possible models for the ozone demand based on data from the pilot study the main objectives were reached. To reach the main objectives they were divided into intermediate objectives:
• Analyze and evaluate the pilot study
• Measurements on the full-scale facility of new parameters
• Perform measurements on the full-scale facility for different ozone doses
• Multivariate analysis to investigate the influence and correlation between different pa-rameters
• Understand the software to be used
• Construct and evaluate models for the ozone demand
1.3
Expected impact of the study
Pharmaceutical residues reach the hydrosphere through the outlet (outgoing water) from WWTPs where negative effects have been shown in the aquatic life. The effects are pre-dominantly a reduced population and a different variety of species, which disrupts the ecosystems. Further, antibiotic resistance might increase in bacteria, resulting in infectious diseases in humans that are difficult to treat. Additionally, the future effects on humans that are exposed to pharmaceutical residues transferred to drinking water sources downstream are unknown. The environmental issues as well as the risk of untreatable diseases arising does not fit with the vision of TVAB. [10, 11]
TVAB works for a comfortable everyday life and a sustainable life cycle for the locals in foremost Linköping and Katrineholm. For instance, they work with waste, recycling, biogas, broadband, electricity-, grid and trading, district heating, remote cooling, drinking water and wastewater. TVAB strives to always be in the forefront of technological development and an example is the pharmaceutical residue treatment from wastewater. A project was initiated in 2014 and with a political interest from Linköping municipality, TVAB decided to investi-gate the possibilities of reducing the loads of pharmaceutical residues from wastewater at Linköpings largest WWTP, Nykvarnsverket. TVAB in collaboration with IVL Swedish Envi-ronmental Research Institute constructed a pilot study with an ozonation step in-between the biological treatment and Moving Bed Biofilm Reactor (MBBR) step in the current process at Nykvarnsverket. The results from the pilot study showed an average reduction of pharma-ceutical residues of about 90 %. Additionally, there was no added toxicity for the aquatic life, no formation of mutagenic by-products and no negative effect on the MBBR-process function . [8]
This pilot study was used as the basis for the design and implementation of the full-scale facil-ity, which is the first permanent, large-scale facility for reduction of pharmaceutical residues in Sweden, inaugurated in September 2017. [8] Optimization, evaluation and verification of the process in full scale, will be performed under the start-up transnational EU-project, Clear Waters from Pharmaceuticals (CWPharma). [12]
1.4
Delimitations
The main constraint was that the full-scale ozonation process could only be operated during the last weeks of this thesis, which altered the initial plan. Therefore, time was lost reformu-lating the purpose. Instead of full-scale measurements on the ozonation step, data from the pilot study in 2014 together with data from new measurements were used to reach the objec-tives. Moreover, there was a limit in the amount of measurements that could be performed and parameters that could be analyzed due to the time restriction of 20 weeks and the re-sources available. Additionally, the time of measurements was restricted by the weather, as similar conditions to those of the pilot study were preferred in order to obtain comparable results and spring was very late this year. Moreover, validation of the model will be limited within the scope of this project. Therefore, it will remain unknown if the model can predict the ozone demand correctly and if the results of the pilot study are valid compared to full-scale data. A more extensive validation will have to be performed later after this thesis has been completed.
2.1
Scientific background
Ozone has several different uses, for example to disinfect drinking waters and to remove pharmaceutical residues from wastewater. Most pharmaceutical residues are excreted in urine and excrements and reach the wastewater treatment plant (WWTP) before they might end up in the hydrosphere where they cause negative effects on the aquatic life and disrupt ecosystems. Thousands of pharmaceuticals are used in Sweden and many of these end up in the WWTPs.[8]
Today, the conventional WWTPs do not have the ability to reduce the majority of the pharma-ceutical residues, including the three pharmapharma-ceuticals in focus during this thesis; diclofenac, metoprolol and oxazepam. [8] With a political interest from Linköping Kommun, TVAB in-vestigated the possibilities to reduce the loads of pharmaceutical residues with ozone to re-ceiving waters (mainly Stångån) at Linköpings largest WWTP, Nykvarnsverket. This was done through a pilot study in 2014.
2.1.1 Ozone
Normally talked about as the gas protecting earth against Ultra Violet-rays, ozone has been used for disinfecting drinking waters for many years. However, recently a new application for ozone, to remove pharmaceutical residues from wastewater, has been of interest. During ozonation several parameters are known to influence the process and the ozone dose required to reduce the pharmaceutical residues present in the wastewater. [5, 13]
Generating ozone
Ozone (O3) is an unstable gas, which must be generated at the point of use. The production
of ozone is performed by an electric current running through oxygen gas (O2), or air, which
leads to the formation of reactive oxygen intermediates. These intermediates then react forming ozone and for every three oxygen, two ozone molecules are generated as follows;
O2+electric energy Ñ 2 O
2 O+2 O2Ñ2 O3
3 O2+energy Ø 2 O3
This reaction is reversible and when the unstable ozone decomposes to oxygen, energy is released. [5, 13]
Reaction mechanism of ozone
Ozone is one of the strongest chemical oxidants and can react either directly or indirectly with a variety of compounds. In the direct reaction ozone reacts with a compound, whilst in the indirect reaction ozone forms radicals which then react with the compound. [13]
Direct reaction
In the direct pathway ozone reacts with an unsaturated bond, due to the un-polar structure, leading to the bond being split. This is known as a Criegee ozonolysis reaction and allows the cleavage of alkene double bonds with ozone. Ozone will react quickly with aromatic compounds like hydroxyl groups, since they carry electron supplying substituents, as well as other electron rich moieties such as tertiary amines and thioethers. [13, 14]
Indirect reaction
The indirect pathway with radicals is very complex, however the reaction can be divided into the three steps; initiation, radical chain and termination. The radicals will react with additional compounds to the direct reaction, such as alkanes and amides. The reaction steps are influenced by substances that act as initiators, promoters or scavengers. Initiators, for example the hydroxide ion, hydrogen peroxide or bivalent iron (Fe2+), are chemicals which initiate the ozone decay and the formation of the radicals. Promoters, for example humic acid, primary alcohols or secondary alcohols, are substances which promote the radical chain. Scavengers, for example hydrogen carbonate, carbonate ion or phosphate, are substances which do not produce radicals required in the radical chain. Therefore, the radical chain is terminated or inhibited. Based on the presence of substances within these groups as well as conditions such as pH or temperature, the direct or indirect reaction will dominate. As the indirect reaction is non-specific and faster than the direct reaction, this will influence how fast the reaction for the compound with ozone is, as well as the degradation efficiency for different substances. [13, 14, 15]
Ozone and Aromatic Compounds
Sometimes ozone reacts directly with an aromatic organic compound, leading to the forma-tion of radicals. As menforma-tioned ozone reacts with electron rich moieties and the radical is formed through an electron transfer from the aromatic organic compound to ozone. As ozone reacts with the electron rich moieties new electron rich sites are generated, leading to radicals being continuously formed when ozone is present. The new sites are generated in the form of phenols, which result either from a direct reaction of ozone with the aromatic organic com-pound or an indirect reaction between the comcom-pound and the radical. The indirect reaction requires the presence of oxygen, which is available during ozonation. [13]
Parameters that influence the ozonation of pharmaceutical residues
During ozonation several parameters influence how much ozone that is required in order to successfully reduce the pharmaceutical residues in the wastewater. Turbidity, Dissolved Or-ganic Carbon (DOC), Chemical Oxygen Demand (COD), nitrite concentration, conductivity (concentration of ions) and Ultra Violet Absorbance (UVA) are amongst the parameters that are known to influence the ozonation process and the ozone demand.
Turbidity
Turbidity, the measure of cloudiness in the water, is measured by applying a light beam to a water sample and measuring the intensity of the scattered light at a 90˝ angle to the light
beam. A higher intensity corresponds to a higher turbidity and a higher turbidity is expected to result in a higher ozone demand. Based on the frequency of the light, different units are used to measure turbidity, for example Formazin Nephelometric Unit (FNU) measured with an infrared light. Turbidity covers several other parameters, both organic and inorganic, such as DOC, COD and nitrite. However, as turbidity depends on the shape, size and refractive index of the particles, observing a direct correlation between turbidity and the weight of the suspended matter in the water is difficult. Therefore it is of interest to measure additional parameters even though turbidity can be used as an indication for an increase in suspended materials and thereby an increased ozone demand. [16, 17]
DOC and COD
Dissolved organic matter, commonly measured as DOC, largely determine the stability of ozone in wastewater, where an increased DOC concentration corresponds to a reduced sta-bility of ozone. The nature of DOC influence the reaction rate and thereby ozone lifetime, which in turn affect the reduction of pharmaceutical residues. This reduction depends on the lifetime of ozone in the wastewater, where a longer contact time between pharmaceutical residues and ozone leads to a better reduction. Ozone reacts with the electron rich moieties of DOC and the reaction leads to the formation of radicals during ozonation. As the reaction with radicals is faster than the direct reaction with ozone the formation of radicals from DOC reduce the stability and lifetime of ozone in the water. This means that an increased DOC concentration shortens the lifetime of ozone in the wastewater, thereby influencing the reduction of pharmaceutical residues and leading to more ozone being required at higher DOC concentrations. [13, 18, 19]
Another parameter which influence the stability of ozone is carbonate alkalinity, which act as a scavenger for radicals. Other parameters, mainly DOC and COD can also act as scavengers. This means that the ozone stability and lifetime decrease and thereby also the reduction efficiency for pharmaceutical residues. [13, 18, 19]
COD can also be used to measure organic content and has been correlated with the ozone demand where a higher COD concentration corresponds to a higher ozone dose require-ment. COD is reduced faster than DOC as the oxidation of an organic compound initially de-crease the COD concentration whilst DOC dede-crease after the organic compounds have been mineralized. Hence, COD concentrations are affected more rapidly by ozonation than DOC concentrations. Both DOC and COD can be determined by adding a strong oxidant, usu-ally dichromate or permanganate, and measuring the amount required for oxidation of the sample. If permanganate is used higher concentrations for COD are determined as this is a stronger oxidant. [13, 18, 19, 20]
Nitrite concentration
Nitrite is generally removed by biological processes in WWTPs. However, if this removal is incomplete ozone reacts quickly with nitrite present in the water. Hence, at a higher nitrite concentration more ozone would be required to oxidize nitrite to nitrate. A high resolution dual beam spectrometer, using UVA, is required to measure the nitrite concentration, where a peak in absorption is observed around 210 nm. However, at this wavelength there is an over-lap with nitrate as well as some organics. This makes on-line measurements of nitrite com-plicated, especially with the low concentrations normally detected in WWTPs of 0.1 mg/L. Therefore, nitrate is used in on-line measurements since there is a known correlation between nitrate and nitrite, where the trends mirror each other. [13, 17]
Conductivity
Inorganic compounds, like ions, can be oxidized by ozone. This means that higher ionic con-centrations correspond to a greater ozone demand. Conductivity can be used to measure various ionic concentrations in one parameter, for example iron, chloride and bromide. The unit for conductivity is usually microsiemens/meter (mS/m). Changes in conductivity nor-mally arise from nitrogen removal in biological treatment. Conductivity is easily determined through measuring the changes in salt concentrations on-line with for example electrodes. [13, 21]
UVA
UVA does not directly influence the ozone demand but can be correlated to other parame-ters such as DOC and COD. A higher UVA corresponds to higher concentrations of organic materials as this usually is composed of aromatic compounds, which absorb UV-light at 254 nm. As higher DOC and COD concentrations correspond to a higher ozone demand there is an indirect correlation where a higher UVA corresponds to a higher ozone consumption. Additionally, UVA has been tested as a control strategy for the ozone dose required to reduce the concentrations of pharmaceutical residues at different conditions. The results showed that UVA, or UV reduction meaning the difference in UVA before and after ozonation, is very efficient as the basis for a regulatory strategy. [9, 13, 17, 18]
Other Parameters
In addition, other parameters, such as pH and temperature influence the ozonation. The for-mation of radicals increase at a higher pH and at values pH>10 only the indirect pathway occurs. The efficiency of a biological treatment step is influenced by temperature, where a higher temperature increase the reduction of organic and inorganic materials. Hence, if a biological step is present before the ozonation a higher temperature would reduce the ozone demand as the organic and inorganic material is reduced. In addition to reducing the con-tent of organic and inorganic compounds in the water, ozonation will reduce pharmaceutical residues in the wastewater, thereby preventing them from reaching the environment. [13, 18, 22]
2.1.2 Pharmaceuticals
Pharmaceutical residues that are released into receiving waters have a negative effect on the environment and the aquatic life. Therefore, a list of substances to be monitored have been compiled for pharmaceutical in different groups such as anti-inflammatory drugs and antide-pressants. [22, 23]
Pharmaceuticals to be monitored
In 2014 a list of 22 pharmaceuticals to be monitored in Sweden was compiled. In addition, EU has a watch list of substances which must be monitored that were included in this list. In this thesis the focus lies on three of the listed substances; diclofenac, metoprolol and oxazepam. [23, 24]
Diclofenac
In Sweden there are national limits for substances that are hard to degrade at WWTPs and diclofenac is the only environmentally hazardous substance which exceeds its limit. As diclofenac is easily accessible, being a non-prescription drug, the use of this substance is very high. Additionally, diclofenac is the active substance in Voltaren, which is fre-quently marketed leading to a further increase in use. Diclofenac is a non-steroidal anti-inflammatory drug which inhibits cyclooxygenase resulting in a reduced prostaglandin for-mation. Prostaglandin cause pain and inflammation and the use of diclofenac reduces these effects. Diclofenac is taken between 1-4 times of the day depending on the form, usually in pills or a gel. As the pharmaceutical is non-prescription and only used when needed there are no recommended dosing intervals or times of day to take diclofenac. The structure of diclofenac, see figure 2.1, contain an ozone reactive phenyl group indicating that ozonation would reduce diclofenac. [25, 26]
Metoprolol
Metoprolol is a β-blocker which binds to β-receptors, thereby inhibiting the hormones adrenaline and noradrenaline from binding and exerting their full effect. Adrenaline and noradrenaline are released into the blood during stress, physical or psychological activity and then bind to β-receptors in the heart and blood vessels, leading to an increase in blood pressure. By inhibiting the binding of adrenaline and noradrenaline to β-receptors the heart rate slows as well as the pumping force thereby reducing the blood pressure and preventing hypertension. Metoprolol is a commonly used β-blocker that is selective for the β-receptors in the heart. Pills with metoprolol are prescribe and usually taken 1-2 times a day, but the dose can be individually adapted. The structure of metoprolol, see figure 2.1, contains both an aliphatic chain and an aromatic ring, which are ozone reactive groups, indicating that the ozonation should be efficient in reducing the metoprolol content in the water. [27, 28] Oxazepam
Oxazepam is an antidepressant which affects the GABA-system in the brain. GABA-steroids are valium like substances that bind to the GABAAreceptor, which is also the target receptor
for alcohol and benzodiazepines, like oxazepam. The binding of GABA-steroids has sedative, anti-depressing and relaxing effects, which are reinforced by oxazepam. As a prescription drug, oxazepam is usually used once a day, but can be taken up to four times a day if needed. The structure of oxazepam, see figure 2.1, lack ozone reactive groups and the aromatic rings are deactivated by electron negative groups, causing a low ozone reactivity. [8, 29]
Figure 2.1:The structure of A) Diclofenac, B) Metoprolol and C) Oxazepam
Effects of pharmaceuticals
Pharmaceuticals residues that are excreted in human urine or feces reach the WWTPs and while some substances are removed by the current process, many pharmaceutical residues are released into receiving waters where they have a negative effect on the environment, since they still have a pharmacological effect. [22]
Aquatic life
The effects of waterborne diclofenac on rainbow trout was investigated and the study showed that the exposure to diclofenac caused tissue damage at concentrations of 1 µg/L. Damages include inflammation, hyperplasia and necrosis in the kidney. Exposure to diclofenac also altered mRNA expression and could have a significant impact on the health of fish long-term [30]. Studies have also shown that diclofenac can cause DNA damage which leads to im-munosuppression as well as genotoxicity in fish [31]. Additionally, diclofenac concentrations lower than 1 µg/L can cause liver damage in rainbow trout [32]. In Stångån a diclofenac concentration of 0.48 µg/L was calculated based on measurements during the pilot study. [8] In a study on zebra-fish embryos metoprolol exposure resulted in scoliosis, heart abnormal-ities and growth retardations at doses of 25 mg/L. However, these effects were observed for metoprolol doses higher than what has been observed in surface water, of 0.2 µg/L and therefore seem like an insignificant risk for fish. Although in more sensitive aquatic organ-isms, green algae and crustaceans, metoprolol exposure at lower doses have shown negative effects, for example an effect on heart rate for crustaceans [32]. Additionally, metoprolol is considered as toxic for aquatic life and negative effects on the growth of green algae have been reported [33]. In Stångån a metoprolol concentration of 3.09 µg/L was calculated based on measurements during the pilot study. [8]
The treated effluent from a Swedish WWTP had an oxazepam concentration of 72 µg/L in a study that investigated the alterations in the behavior of perch after oxazepam exposure. The study showed a bioaccumulation of oxazepam in muscle tissue and significant effects
on the behavior of perch. Treatment with a low dose of oxazepam, 1.8 µg/L, resulted in asocial and more active individuals, compared to the untreated fish. Additionally, oxazepam increased the feeding rate of fish which might have effects on the ecosystem in the future. [34] A oxazepam concentration in Stångån of 0.3 µg/L was calculated based on measurements during the pilot study. [8]
Environmental effects of sex hormones on the aquatic life are well known, for example the feminization of male frogs and fishes by the hormones, ethinyl estradiol, estradiol or a gestrel, in the contraceptive pill [31, 35]. A study reported that the exposure to levonorgestrel at an early stage in life of frogs had several effects that manifested themselves in adults. Female frogs exposed to levonorgestrel throughout life became sterile, as a result of severe disrup-tion in the ovary and oviduct development. Addidisrup-tionally, levonorgestrel has been shown to bioaccumulate in fish and exposure resulted in inhibition of the egg-laying for female adult fish [36]. In a study of zebrafish exposed to progesterone or norgestrel a disruption in sex differentiation was observed. For progesteron the proportion of females increased, at a dose of 63 ng/L, while an increase in the proportion of males was observed for norgestrel at 34 and 77 ng/L. It is thought that the exposure to these compounds alter the transcriptions of genes related to the synthesis of sex hormones and thereby the levels of sex hormones in zebrafish. [37] Levonorgestrel was calculated to a concentration of <0.432 µg/L in Stångån based on measurements during the pilot study. [8]
Resistance in bacteria
There is a concern that bacteria resistant to antibiotics are formed in WWTPs and reciving waters as well as a concern that the effluents from WWTPs contain antibiotic concentra-tions close to effect levels. Resistant bacteria have been found in sludge from the wastewater treatment, these bacteria were resistant to various antibiotics at high levels and the number of resistant bacteria increased in the summer. Additionally, increased resistance in bacte-ria after long-term exposure to subtherapeutic antibiotic concentrations have been reported. Also, microorganisms exposed to antibiotics that become resistant can transfer their genes to pathogenic bacteria, which can result in infections in humans that are difficult to treat. [10, 31]
Other organisms
When vultures were found to be endangered due to the consumption of diclofenac from their food, dead cows treated with diclofenac in India, concerns about pharmaceuticals in the environment arose. The decline in vulture population with 95 % in the 90’s resulted from renal failure and visceral gout, which were the effects of diclofenac consumption. The source of diclofenac might also be from the water source, although the concentrations of diclofenac in water is low and bioaccumulation does not occur in vultures, making this a negligible source of diclofenac compared to the dead cows. [38]
Humans consume pharmaceutical residues through water and food, mainly fish at low con-centrations and so far, no effects have been observed. However, effects of pharmaceutical residues on humans and animals have not been investigated for a longer time period. Also, the exposure to multiple pharmaceuticals have not been investigated. Additionally, the fre-quent introduction of new drugs results in their being substances with effects that are yet to be determined. Moreover, the effects of exposure to pharmaceutical residue might have a different impact due to individual variations in sensitivity as well as the sensitivity of certain groups, for example the elderly. [22]
2.1.3 Wastewater treatment
The water from the sewage system, industries as well as surface water must be purified to avoid negative effects on the environment. This is performed by WWTPs, using mechanical, biological and chemical treatment techniques. In Linköping, TVAB is responsible for the wastewater treatment and for making sure that water being returned to the environment meets the set requirements. [5, 39, 40]
Nykvarnsverket
The WWTP of TVAB in Linköping, Nykvarnsverket, was built in 1952 and has since under-gone several changes, the latest being the addition of an ozonation step, for pharmaceutical residue treatment, in 2017. Today, Nykvarnsverket meets the requirements for water being returned to the environment and with the ozonation this will be the case in the future as well. In the process, see figure 2.2, the outgoing water is not the only product from the WWTP. Additionally, biogas used in buses for public transport and sewage sludge used as fertilizer in agriculture, is produced at Nykvarnsverket. However, the focus in this report is the wastewater side of the process and especially the ozonation step. [39, 40]
Untreated wastewater poses several issues, mainly littering the recipient, spreading disease, eutrophication of waters, spreading environmentally hazardous substances and the latest concern with the effects of pharmaceutical residues on aquatic life. This is prevented by the wastewater treatment process at Nykvarnsverket, where the main steps are; [5, 39, 40]
1. Screens 2. Grit chamber 3. Aeration 4. Primary clarifier 5. Biological treatment 6. Ozonation 7. MBBR
8. Secondary clarifier/chemical treatment
The wastewater treatment process starts in a general way with mechanical treatment using the first two steps; screens and grit chamber. The screens are used as a first step to remove larger objects, such as plastics, toilet paper and ear swabs preventing the littering of the re-ceiving water, Stångån. The water then proceeds to the grit chamber, where heavier particles such as sand and coffee-grounds sediment. In the inlet to the grit chamber iron sulfate is added as a precipitation chemical. The iron is oxidized, from bivalent to trivalent, in the grit chamber, as well as in the following aeration step, leading to the formation of flocs. These flocs are composed of the iron, phosphorus and organic material, which sediment in the fourth step, the primary clarifier. Cat- and anionic polymers are added before this step to enhance the effect of the iron and to ensure that phosphor is removed from the water as this is a major source in eutrophication. The sedimented flocs form sludge, which is then used for biogas production. [5, 39, 40]
The biological treatment is the fifth step, where organic materials and nitrogen are removed by microorganisms. The aeration in this step is intermittent, making the environment aerobic and anoxic in relation to when the aeration occurred. The organic material is either degraded to carbon dioxide and water or converted into new biomass. Additionally, remaining iron
is oxidized, form flocs and sediment in the biological treatment and the sludge formed in this step is returned back into this step to ensure that the microorganisms remain in the process. Also, nitrogen is converted to nitrogen gas in two steps by the microorganisms in the biological treatment step. Nitrification is the first step, where nitrogen in the form of ammonium is oxidized to both nitrate and nitrite by bacteria in an aerobic environment. Then denitrification, where the both nitrate and nitrite are oxidized to nitrogen gas by other microorganisms in an anoxic environment, is the second step. The nitrogen gas is released into the atmosphere, which already has a high nitrogen content of almost 80 %. [5, 39, 40] After the biological treatment the newest process step, which will be covered more in detail below, is to take place. The pharmaceutical residue treatment with ozone will remove an-tibiotics, antidepressants, painkillers as well as bacteria resistant to antibiotics once running. This is the first permanent full-scale facility in Sweden and was inaugurated in the fall of 2017. [39, 40]
After ozonation is the MBBR step, with carriers coated in biofilm. In this step nitrogen is con-verted to nitrogen gas in two steps by the microorganisms in the biofilm, like in the biological treatment step. Nitrification and denitrification occur in separate tanks which are aerobic and anoxic, respectively. Ethanol as a source of carbon and phosphoric acid as a source of nutrition for the microorganisms in the biofilm are added between the nitrification and deni-trification. In the last step, the secondary clarifier or chemical treatment, aluminum chloride is added to precipitate phosphorus, which is done as before with the formation of flocs which then sediment and form sludge. After this step the water exits the process and goes into the recipient, Stångån. [5, 39, 40]
Figure 2.2:The wastewater treatment process at Nykvarnsverket.
Pharmaceutical residue treatment
The first permanent full-scale ozonation facility for continuous use was constructed in 2017 based on successful results of a pilot study at Nykvarnsverket in 2014, see section 2.1 below. The goal with adding ozonation in the process is to achieve a 90 % reduction of pharmaceuti-cal residues in the outgoing water, thereby preventing the negative environmental effects.[39, 40] Ozone is generated by running an electric current through liquid oxygen and the ozone injector as well as the water basin have been designed to ensure an effective mass transfer, see figure 2.3. A high Mass Transfer Efficiency (MTE) is obtained by having an efficient mixing of ozone with water. Part of the biologically treated water from the previous step is mixed with ozone, whilst the remaining water enter the basin from the bottom. In the injector, placed in the top of the water basin, the speed of the incoming water is increased, creating a
vacuum. This vacuum results in the ozone gas being drawn into the water. As the water and ozone exits the injector the speed is reduced and mixing blades create ozone microbubbles. Since these bubbles increase the contact surface between ozonated water and the biologically treated water, the mass transfer becomes more efficient. Additionally, a radial diffuser placed at the outlet of the injector is used to distribute the water-ozone mixture efficiently and also to create more microbubbles. This gives a larger contact surface which is prerequisite for an efficient mass transfer of ozone from the gas- and water phase. By having a large contact surface the process can operate more effectively, as ozone will react with the water that it comes into contact with.
The water basin was designed to ensure that the water moves homogeneously through the reactor so that no dead zones or short circuit streams forms. Additionally, the contact time between ozone and water is evenly distributed in the water volume with the design, result-ing in more efficient reduction of pharmaceutical residues in the water. Both on-line and off-line measurements can be made at certain sampling points to monitor the process. The ozonation was placed after the biological treatment to avoid that ozone reacts with organic materials, which can easily and more cheaply be degraded by the microorganisms. After the pharmaceutical residue treatment, the MBBR-step has been shown to be an efficient method for removing by-products formed by the ozonation. Hence, the ozonation was placed here, making it the sixth of eight steps in the process at Nykvarnsverket. [5, 39, 40, 41]
Figure 2.3: The ozonation process with ozone generation from liquid oxygen and on-line measurement points (On 1, On 2 and On 3).
2.1.4 Pilot study
With a political interest from Linköping Kommun, TVAB investigated the possibilities to re-duce the loads of pharmaceutical residues to receiving waters (mainly Stångån) at Linköpings largest WWTP, Nykvarnsverket. This was done through a pilot study in 2014.
Background
Over a thousand different pharmaceuticals with different active substances are used in Swe-den. These pharmaceutical residues reach the WWTPs via urine and excrement and might then end up in the hydrosphere. Analgesic and anti-inflammatory pharmaceuticals like paracetamol, ibuprofen, naproxen and diclofenac are the dominating pharmaceuticals in the influent. However, some of these pharmaceuticals, especially paracetamol and ibuprofen, are reduced effectively in the conventional WWTPs today. In the outgoing water it has been ob-served that furosemide, metoprolol, atenolol and diclofenac are not reduced in the WWTPs. Additionally, pharmaceuticals for the central nerve system, like the substance oxazepam, have a limited reduction in the WWTP.
Current studies have shown no effect of the pharmaceutical residues on humans. How-ever, it has been shown that the pharmaceutical residues in low concentrations (ng-µg/L) have negative effects on the aquatic life and environment, for example causing sterility and personality disorder in fish. Moreover, the microbial ecosystem may be disrupted, which in turn may affect higher ecosystems. Although the debate regarding the pharmaceuti-cal residues was new and knowledge limited, the politicians in Linköpings Kommun had an interest to investigate whether a reduction of pharmaceutical residues in Linköpings wastewater was possible or not. TVAB in collaboration with IVL Swedish Environmental Research Institute then decided to construct a pilot study at the biggest WWTP in Linköping, Nykvarnsverket. The goal with the pilot study was to investigate whether ozonation would be a possible method to use in order to reduce the pharmaceutical residues in the wastewater. Ozonation and adsorption with active carbon as purification techniques were initially of in-terest for TVAB. Both techniques have a relatively similar pharmaceutical residues reduction capacity, but an evaluation witch focus on ecotoxicity, nitrogen-, phosphorus-, Biochemical Oxygen Demand (BOD) reduction and bacterial reduction as well as energy costs resulted in that the ozonation technique was deemed to have the most beneficial prerequisites and to be the best fit for TVAB.
Before the ozonation process was installed, the contents of pharmaceutical residues was mea-sured after the biological treatment process, which gave a good indication of the amount of pharmaceutical residues that end up in the recipient. TVAB found 28 substances, which was listed in a priority list with different levels of risk (high risk, moderate risk and low risk). There were 5 substances with high risk in the priority list, including oxazepam and metopro-lol. Diclofenac on the other hand, was assessed as a moderate risk. [8]
Method
The pilot plant was set-up at Nykvarnsverket in Linköping and the pilot study was carried out over one year, where the ozonation trial period was about 5 months including a control trial and a repeat trial. The ozonation trial period included several studies and measure-ments, including daily variation mapping, ecotoxicological- and dose response studies for ozonation.
The reaction tanks were two series bubble columns with a volume of 0.115 m3each, where ozone was added from the bottom of the columns. 12 different doses of ozone were tested
(where 1.8 mg/L was the lowest ozone dose and 23.1 mg/L the highest ozone dose). The retention time was held constant at 11 minutes and the flow was 1.5 m3/h for all doses tested, with the exception of the two lowest ozone doses, where the flow was set to 1.7 m3/h. During the trial period, samples were taken before, during and after ozonation. Additionally, samples were taken during and after the MBBR step in order to analyze the by-product forma-tion and to evaluate how the ozonaforma-tion step affects the process in general. The samples were analyzed in laboratories, both at TVAB but also at IVL Svenska Miljöinstitutet, Swedish Uni-versity of Agricultural Sciences (SLU) and Toxicon AB. Furthermore, parameters like UVA, temperature, pH, flow, nitrate and turbidity were measured on-line. [8]
Results
In summary, TVAB observed a significant reduction of the pharmaceutical residues when the ozonation process was operated. Toxicity for the aquatic life did not increased and no formation of mutagenic by-products or increased gene toxicity could be observed. Moreover, there were no negative effect on the denitrification in the MBBR-step caused by the ozonation. Daily variation of the water flow into the WWTP as well as daily variation of the pharma-ceuticals were observed. For some groups of pharmapharma-ceuticals, like antibiotics, which one typically take twice a day, there was a dilution during the morning due to the increased water flow. Comparable, for pharmaceuticals which one typically take once a day, like an-tidepressant pharmaceuticals, no variation in concentration could be seen. Moreover, it was observed that the flow was lower during the summer period, probably due to the vacation when many students temporary move and several big industries in Linköping shut down. This means that a seasonal variation in flow occur as well.
The daily and seasonal variation of the flow in to the WTTP affected the ozone residual, i.e. remaining ozone in gas phase. A higher flow of the water gave a shorter retention time in the biological treatment step. Furthermore, the shorter retention time resulted in an increase of organic materials in the water, hence more ozone was consumed which lead to a decrease in the ozone residual.
A correlation between reduction of pharmaceutical residues and ozone sensitive UVA was observed, i.e. a higher reduction of pharmaceutical residues resulted in a lower UVA. On the other hand, there was a lack of data and uncertainties in analyzes during the pilot study. It can therefore be beneficial to evaluate UVA in correlation with pharmaceutical residues further.
Nitrite concentration was another parameter which was of interest during the pilot study. Nitrite is oxidized to nitrate in presence of ozone, which means that nitrite concentration directly affect the consumption of ozone. The highest and the lowest nitrite concentration differed with 0.5 mg Nitrite Nitrogen(NO2-N)/L, which corresponded 1.7 mg/L increased
ozone demand.
All pharmaceuticals measured were reduced in the ozonation process. After the ozonation (with a ozone dose of 5 mg/L), only one of the substances, oxazepam, was assessed as a high risk. Metoprolol on the other hand, end up as a moderate risk. Diclofenac had a risk factor lower than 0.01 and was therefore removed from the priority list post ozonation. Based on the result, an ozone dose of around 5-8 mg/L depending on externals seems preferable for TVAB in order to reach 90% reduction of pharmaceutical residues which would minimize the risk of negative effects on the environment. [8]
Possible Improvements
Due to the results of this pilot study, it is obvious that the ozonation is a complex process with many parameters to take in consideration. The wastewaters differ among the WWTPs and there are consequently no general guidelines to follow. Underdosing may result in insufficient pharmaceutical residues elimination while overdosing of ozone may result in more oxidation products and unnecessary expenses. It is therefore important for TVAB to find the ideal dosage of ozone, preferable using a model and/or automatic control strategies. [9]
Moreover, TVAB together with 14 other organizations, participates in an EU-project called CWPharma, where the purpose is to reduce the active pharmaceutical substances in the Baltic Sea. With financing from EU, the contributions have the opportunity to test and evaluate purification techniques that reduce the pharmaceutical residues from aquatic environment. In TVABs case, they have the opportunity to optimize the ozonation process at Nykvarnsver-ket, Linköping. [12]
Constructing a model or control strategies to the full scale-process from the historic data of the pilot study can bring difficulties due to the difference in conditions or constructions. In this case, the ozone reactor in the pilot facility had a size of 2 x 0.115 m3comparable with the 600 m3in the full-scale facility. Moreover, the flow was held constant in to the ozonation step during the pilot trial period, which is not the case in the full-scale facility. Another aspect to take in consideration is the population growth in Linköping during the last four years of approximately 7000 inhabitants. [42]
2.2
Methodology
A timeline illustrating how the work was carried out can be seen in figure 2.4. The initial phases during this thesis were a pre-study and a literature study. This was followed by the multivariate analysis and model construction based on different data from the pilot study. Measurements were performed on the wastewater in the full-scale facility in order to verify the results from the pilot study. Moreover, measurements were performed to find new ozone consuming parameters. Further, the different models that were constructed in the multivari-ate analysis were evalumultivari-ated in order to find the best one. The chosen model was then tested in MATLAB with a test file, at the same time as the pilot study was verified.
Figure 2.4:A timeline illustrating how the work was carried out.
2.2.1 Pre-study
During the pre-study the focus was on planning the project by setting goals, determining activities to be performed in order to reach the goals and allocating time for each activity. Information was gathered by talking to internal as well as external stakeholders. Internally, meetings with supervisor and the laboratory at TVAB were held in order to decide which measurements that were of interest, but also when they should be performed. Moreover, meetings with external stakeholders around north Europe (Sweden, Germany and Denmark) were held to exchange knowledge and experiences within ozonation of wastewater. Ad-ditionally, a participation at the NAM-conference (Nationella Konferensen Avlopp & Miljö) gave interesting inputs and ideas to this thesis. The result from the pre-study was a planning report, which contained the objectives, goals, boundary conditions, methods, activities, mile-stones, a gantt chart and a short theoretical background. The plans in this report has since been revised with the original plan and goals in mind.
2.2.2 Literature study
To receive information about ozonation in WWTPs as well as different pharmaceuticals, a lit-erature study was carried out. Additionally, the pilot study was thoroughly studied in order to get a deeper understanding of the conditions for Nykvarnsverket. Moreover, the achieved results from the pilot study were essential to complete the main goals in this thesis. The pilot study results were compiled in a report, which was received from TVAB. The aim with the literature studies was mainly to find information about the parameters that were found to be interesting in the pre-study. Additionally, it was of interest to find other WWTPs that had done similar projects. Literature received from Linköpings University library database and TVABs internal system were the most used where ebooks, reports and articles were of interest. The outcome of the literature study was an understanding of how the conditions at different WWTPs affect the ozonation process and that there is no standard operating procedure. [13,
18, 43] Moreover, the degrading ability for the pharmaceuticals of interest (diclofenac, meto-prolol, oxazepam) were investigated and the phenyl groups, aliphatic chain and aromatic rings were found to be the reactive groups within the structures. However, oxazepam has no ozone reactive groups and the aromatic rings are deactivated by electron negative groups, making oxazepam hard to reduce from the wastewater. [8, 13]
2.2.3 Main activities
The knowledge and results from the pre-study and the literature study gave an idea about which parameters that could be of interest when the main activities were planned. However, the plans were revised during the project since problems with the ozonation arose. To reach the goals for this thesis, a multivariate analysis and measurements on the full-scale facility were decided to be the main activities. The multivariate analyzes were performed to inves-tigate the influence and correlation of various parameters that were found to be of interest during the literature study. To verify the results from the pilot study, new measurements were made on the full-scale ozonation.
Multivariate analysis and regression
Before the multivariate analysis, the data from the pilot study was sorted. Days when it was heavy rain, problems with the equipment due to a thunderstorm or a power outage were excluded. Moreover, the data was sorted based on dose-, control-, and repeat trials. All data used was an hourly average of the standard data. The data for the regression fitted line plot was based on data from an ideal day, meaning when the behavior of nitrite and the ozone residual was as expected according to literature. [13]
The multivariate analyzes were conducted in Minitab, which was received from Linköping University. Minitab was selected as it is a user friendly and easily manageable statistical software [44]. Several functions were used, like Principle Component Analysis (PCA), Partial Least Square (PLS) and Regression - fitted line plot. The PCA was used in order to identify outliers in the data. PLS was used to identify which predictors (parameters) that correlated to the response variable (ozone residual), but also how much the parameters correlated to each other and the response variable. [44, 45] The Regression function was used in order to fit data to a cubic regression line.
The results from the multivariate analysis were three different models which were evaluated and the best model was tested in MATLAB.
Measurements on full-scale facility
Measurements were performed on the full-scale facility in order to verify the pilot study and find new parameters that affect the ozone consumption and thereby the ozone demand. To verify the pilot study, the conditions needed to be similar to those of the pilot study, meaning during spring temperatures and no heavy rain or snow melting. The amount of pharmaceutical residues before and after the ozonation process were investigated to confirm that all of the substances were reduced to 90% in the full-scale facility. Even though it was only possible to measure daily average on other parameters like DOC and suspended solids, these results was compared to the results from the pilot study. The effects on these parame-ters from the different conditions in the full-scale facility compared to the conditions during the pilot study were investigated.
Moreover, new interesting parameters like COD, Fe2+ and conductivity were measured.
see whether these parameters correspond to a part of the ozone consumption. Unfortunately, these parameters could not be included in the model construction due to the lack of data. However, the results may be valuable for TVAB in the future.
Depending on which parameter that was measured, the different samples were send to dif-ferent laboratories. The sample for COD analysis was sent and analyzed by SYNLAB in Linköping. Moreover, the pharmaceutical residues were sent and analyzed by a laboratory at Aarhus University, Denmark. The rest of the samples were analyzed by the laboratory at TVAB. The results were compared with the result from the pilot study in order to verify the ozonation process and thereby the pilot study.
2.2.4 Validation
An important part in this thesis was to find a model that could predict the ozone demand based on different conditions in the wastewater. Alternative models that could predict the ozone residual were constructed in order to find the best one. The models were validated by comparing real data with data received from the models and the best model was selected. The data used was, as mentioned above, from the pilot study due to lack of data from the full-scale facility. The result from the validation was the selection of the model which fitted the validation data best.
2.2.5 Verification and testing
The pilot study was verified by comparing results from the new measurements with pilot study results. In parallel, the chosen model was then tested in MATLAB with a test file. Verification
To verify the pilot study, the new results were compared to the results from the period when the pilot study was carried out. Unfortunately, the ozonation could only be run for approx-imately three hours before a filter was clogged. The comparison was therefore limited and only few of the parameters could be included in the comparison. On the other hand, mea-surements of the pharmaceutical residues could be done and thereby it was possible to see if the ozonation process in the full-scale facility worked as expected. The reduction of phar-maceutical residues back in 2014, during the pilot study, was compared with the reduction during the spring of 2018. The results from this made it possible to verify the pilot study and the full-scale facility.
Testing
The best model was tested in MATLAB to confirm and evaluate the model. A script was constructed in order to calculate different ozone doses depending on different conditions. A test file with different conditions for the parameters included in the model was run. The ozone dose was set to have a lower limit of 4 mg/L and a higher limit of 8 mg/L in order to avoid the consequences for over-, and underdosing ozone. [9] Moreover, a start dose was set to 6 mg/L, right in between these two limits.
2.3
Models
For the multivariate analyzes both PCA and PLS were used.
2.3.1 PCA
PCA is an analysis where the aim is to identify a smaller number of uncorrelated variables. However, in this thesis, a PCA was used in order to identify outliers in the data. Points over a calculated reference line were classified as outliers and thereby excluded in further analysis. The outliers were calculated and visualized in the software (minitab), specified in Mahalanobis distance (MD) which is the most common way to identify outliers. The MD between two objects was defined as follow [46];
d(Mahalanobis) = b (xi´¯x)1ˆ(xi´¯x) C , where $ ’ & ’ % xi =an object vector
¯x=arithmetic mean vector C=sample covariance matrix
2.3.2 PLS
PLS is a statistical technique where it is possible to investigate the relationship between a response variable and different predictors in a multivariate dataset [45]. The algorithm that is used is a nonlinear iterative partial least square (NIPALS) algorithm. The number of predictors is reduces with a technique that extract a set of components, which describes the maximum correlation between the response variable(s) and the predictors. The components are selected depending on how much variance they explain in the predictors, as well as the variance between predictors and response variable(s) [44];
Two matrices, X
(nˆp)and Y(nˆq)are assumed to be in linear decomposition like follow:
X=TP1+E , where $ ’ & ’ % T=x-score P1 =x-loading E=x-residual Y=UQ1+F , where $ ’ & ’ % U=y-score Q1=y-loading F=y-residual
PLS extract factors from X and Y such that covariance between the extracted factors is max-imized and x-, and y-scores are received. Each extracted x-, and y-score are linear combina-tions of X and Y respectively. Every x-, and y-score gets an eigenvalue and from these, U can be estimated and thereby Y predicted. However, a function for the linear predictable model can be received based on the X and Y matrices.
The model is in the form of Y=XB+N , where $ ’ ’ ’ ’ & ’ ’ ’ ’ % Y=response matrix X=predictors matrix
B=regression coefficient matrix N=noise term
The different coefficients for all predictors are calculated from the linear regression line. The value from the regression line is subtracted from the real value, the difference between these two values are the error in the data. Hence, the coefficients are calculated as the square of sums of error;
n
ř
i=1
3.1
Method and materials
Data for parameters that were measured during the pilot study was available, see table 3.1. Additionally, new measurements were made for parameters that were measured during the pilot study as well as for new parameters. Samples were taken at different points in the ozonation process and analyzed by different parties, the laboratory at TVAB, SYNLAB and Aarhus University.
Table 3.1:Available data from measurements made and also measurements and on-line data from the pilot study.
New Data Historic Data (Laboratory) Historic Data (on-line) COD (dissolved + total) Nitrite Turbidity
Conductivity DOC Temperature
Fe2+ Total Organic Carbon pH
DOC Suspended solids Flow
UVA UVA Nitrate
Nitrite Gas flow (ozone injected)
Off-gas ozone (ozone residual)
3.1.1 Data analysis
The extensive on-line data of 43 557 data points for several parameters from the pilot study was first structured based on days when the same concentration of ozone was used and also divided based on different gas flows. The on-line data was taken in intervals of six minutes and to reduce the amount of data as well as make it possible to include additional parame-ters, hourly and daily averages were calculated for the days of interest. The daily averages calculated for the days of interest were used to investigate the influence of an altered gas flow on the ozone residual.
