Water Quality
The influent wastewater water quality over the duration of the sampling period is shown in Table 1. UVT254 ranged from 39.7 to 70.6% (mean = 61.8%) and UVT280 ranged from 44.7 to 75.7% (mean = 66.8%) and flow ranged from 164 to 1010 m3day-1 (mean = 490 m3day-1). These data indicate that the water quality and flow was variable over the sampling period and further strengthens the disinfection data because a variety of wastewater conditions were captured. The TSS was observed to be very low for a wastewater facility with a mean value of 5.5 mgL-1 and a max value of 9.5 mgL-1. Total iron concentrations were also found to be low with a mean value of 0.21 mgL-1 and a max value of 0.36 mgL-1. Low TSS and iron concentration suggest that particulate matter has a limited impact on disinfection performance.
Table 1 Mean, minimum, and maximum values for select water quality parameter from duration of study
Parameter
|
Mean (95% CI)
|
Min
|
Max
|
n
|
Flow, m3day-1
|
490 (361 - 618)
|
164
|
1009
|
15
|
Intensity, mWcm-2
|
27 (19.2 - 34.8)
|
11.1
|
55.1
|
15
|
Total Iron, mgL-1
|
0.21 (0.16 - 0.25)
|
0.09
|
0.36
|
16
|
TSS, mgL-1
|
5.5 (4.2 - 6.9)
|
1.0
|
9.5
|
16
|
UVT254, %
|
61.8 (58 - 65.6)
|
39.7
|
70.6
|
16
|
UVT280, %
|
66.8 (63 - 70.6)
|
44.7
|
75.7
|
16
|
Full-scale plant data were also collected as part of the typical sampling for the duration of the study period. Operators collected flow, influent and effluent TSS, influent and effluent pH and effluent E. coli concentrations approximately every 2 weeks. Table 2 summarizes the parameters relevant to this study. Comparing the full-scale data to the bench-scale data indicates that the range of flows and wastewater quality captured during the sampling period captured the range of typical flows for the full-scale facility. Average flows at the plant over the duration of the study were 471 m3day-1 (average for sampling events = 490 m3day-1). The average TSS at the facility was 6.8 mgL-1 which was a slightly higher than the 5.5 mgL-1 observed in the lab.
Table 2 Full-scale quality data during experiment sampling period
Sample
|
Daily Flow, m3day-1
|
Effluent TSS, mgL-1
|
Effluent E. coli, CFU-100ml-1
|
1
|
632
|
8.6
|
70
|
2
|
638
|
11.6
|
<10
|
3
|
379
|
4.4
|
<10
|
4
|
312
|
10.2
|
<10
|
5
|
415
|
1.0
|
<10
|
6
|
446
|
5.0
|
30
|
Treatment Performance
Figure 1 shows the disinfection performance for the wastewater treatment facility (WWTF) for both LED and LP collimated beam light sources. The plant performance is shown by the grey bar and dashed line for comparison to the bench-scale treatments indicated by the different colored box plots. The design fluence for the reactor installed at the WWTF was 30 mJcm-2, and these results show that UV LEDs at 280 nm outperform the LP at this fluence. Furthermore, the overlap of the plant performance shaded region and the LP bench-scale treatment at the 30 and 40 mJcm-2 in Figure 1 indicate that the system was matrix limited when considering the UV auditing methodology, (i.e. the system is treating the wastewater at the highest quality)19. This was not unexpected as the average daily flows experienced by the facility are only 36% of the design flows. The LED light source outperformed the LP collimated beam at each of the examined fluences at the bench-scale. This result suggests that UV LED light sources are a better tool for disinfection under some wastewater quality conditions.
Collimated beam results for the LP light source indicated that it only achieved disinfection comparable to full-scale LP performance at a fluence greater than 40 mJcm-2. The design fluence of the WWTF is 30 mJcm-2 which suggests that the facility is performing above the design rate. This is not surprising as the average flowrate experienced at the facility was 490 m3day-1 over the course of the study compared to the design flow of 1363 m3day-1. These data show that a significant amount of energy is wasted by the LP system due to an excessive applied UV fluence. A full-scale LED system installed at this WWTF could be better tuned to the changing water quality at this location.
Inactivation Kinetics Modelling
Modelling of each of the disinfection light sources further indicated that there were significant differences in the behaviour of disinfection between the LP and LED treatments. LED light sources reached a disinfection limit at a fluence of 20 mJcm-2 (Figure 2). The LP modelling indicated that the shouldering of the model began at 40 mJcm-2 and would reach steady state at a fluence that was beyond the range of fluences examined in this study.
Table 3 shows the kinetics data and Geeraerd’s model fit for each of the light sources and WWTF data. The effectiveness of the LED versus LP was observed to be significantly different. The 280 LED was found to have a k value that was twice that of the LP system. Practically this means that the 280 nm UV LED requires half the fluence to achieve the same log reduction in E. coli. Moreover, the 280 nm UV LED Nres, or the upper level of treatment, was significantly greater compared to the LP system (3.61 log versus 2.82 log). As this upper limit of disinfection is typically due to particle shielding effects, this suggests that the 280 nm UV LED had a higher propensity to reach bacterial communities that may have attached to the particulate matter in the matrix. Particle shielding effects have been observed to be wavelength dependent, as particle UV absorbance capability increases as the wavelength decreases, which lowers the inactivation capabilities at those lower wavelengths.25. Furthermore, self aggregation of E. coli has also been shown to be wavelength dependent 26,27
Table 3 Kinetics data for Geeraerd’s model fit parameter and Plan LRV for 2 UV Lamps (bracketed values indicate a 95% confidence interval about the mean)
The Nres confidence intervals for each bench-scale light source and WWTF data overlapped and there was no significant difference between the bench-scale treatments and full-scale disinfection performance. This is the first instance where the auditing process captured a plant that was substantially overdosing UV radiation. This result indicates that the UV auditing process improves operational efficiency even for a plant which is operating under ideal disinfection outcomes.
Energy Implications
The increase in the upper level of treatment (+0.79 log) observed for the UV LED source suggests that the interaction of the wavelengths and particulate matter may be influencing performance. Nonetheless, the 33% improved germicidal efficiency when compared to traditional UV light sources begins to address the current discrepancies in wall plug efficiency (WPE) between the two technologies. As of 2020, the highest WPE achieved for a commercially available 280 nm UV LED was 4.1% (LP lamps 30-35%) and equivalent quantum efficiency (EQE) was 6.1%28. Currently, UV LEDs in the 280 nm ± 5 nm range have an EQE ranging from 9-20.3%7,29 and best LEDs would typically be around 7.1%. This marked improvement in the last few years, and forecasted improvements for UV LED WPE indicate that the energy efficiency discrepancy will decrease as LED light sources improve 30,31. Further efficiency can be found through creative design of UV LED reactors such as highly reflective internal surfaces and overall shape of the reactor which allows for maximum interaction of emitted light and target pathogen. These efficiencies combined can improve the feasibility of full-scale implementation of UV LED reactors.
UV LEDs achieved similar disinfection performance to the full-scale WWTF at a UV LED fluence of 20 mJcm-2 whereas the full-scale design fluence was 30 mJcm-2 (Figure 3). It has been illustrated above that the full-scale LP installation was operating to deliver a fluence more than 40 mJcm-2 reduction equivalent fluence (REF), despite a design fluence of 30 mJcm-2, and hence is consuming additional power to treat above the required level. It was also shown that an equivalent level of inactivation could be achieved by an LED system operating to deliver a 20 mJcm-2 REF. A detailed analysis of the energy cost comparison of an equivalent LED system installed at the Springfield Lake site is beyond the scope of this paper, though a baseline comparison may be drawn. Therefore, the fluence to be delivered by an equivalent LED system would be 1.5 – 2.0 times lower than that of the current LP installation.
Energy Implications
The increase in the upper level of treatment (+0.79 log) observed for the UV LED source suggests that the interaction of the wavelengths and particulate matter may be influencing performance. Nonetheless, the 33% improved germicidal efficiency when compared to traditional UV light sources begins to address the current discrepancies in wall plug efficiency (WPE) between the two technologies. As of 2020, the highest WPE achieved for a commercially available 280 nm UV LED was 4.1% (LP lamps 30-35%) and equivalent quantum efficiency (EQE) was 6.1%28. Currently, UV LEDs in the 280 nm ± 5 nm range have an EQE ranging from 9-20.3%7,29 and best LEDs would typically be around 7.1%. This marked improvement in the last few years, and forecasted improvements for UV LED WPE indicate that the energy efficiency discrepancy will decrease as LED light sources improve 30,31. Further efficiency can be found through creative design of UV LED reactors such as highly reflective internal surfaces and overall shape of the reactor which allows for maximum interaction of emitted light and target pathogen. These efficiencies combined can improve the feasibility of full-scale implementation of UV LED reactors.
UV LEDs achieved similar disinfection performance to the full-scale WWTF at a UV LED fluence of 20 mJcm-2 whereas the full-scale design fluence was 30 mJcm-2 (Figure 3). It has been illustrated above that the full-scale LP installation was operating to deliver a fluence more than 40 mJcm-2 reduction equivalent fluence (REF), despite a design fluence of 30 mJcm-2, and hence is consuming additional power to treat above the required level. It was also shown that an equivalent level of inactivation could be achieved by an LED system operating to deliver a 20 mJcm-2 REF. A detailed analysis of the energy cost comparison of an equivalent LED system installed at the Springfield Lake site is beyond the scope of this paper, though a baseline comparison may be drawn. Therefore, the fluence to be delivered by an equivalent LED system would be 1.5 – 2.0 times lower than that of the current LP installation.
Using flowrate as a measure of how large a UV system must be to deliver a 30 mJcm-2 dose an estimation of current energy consumption for UV treatment systems was completed. Using the Springfield Lake treatment facility with a baseline flow of 1363 m3day-1 and an annual energy usage of 5781.6 kWh estimations of annual power consumption across the provincial totals were calculated. Further, using the provincial energy profiles provided by the Government of Canada, the annual CO2(e) generated was estimated36. These values were then used to assess the scenario where all facilities switch to 20% WPE efficient UV LED systems. This analysis indicates that an annual reduction of 946 tonnes of CO2(e) is attainable by operating UV LED disinfection systems (Table 4).
Table 4 Estimated CO2(e) generation from wastewater treatment facilities using UV treatment based on 2022 provincial rates of CO2(e)generated/ kWh and total daily flows.
Province
|
CO2(e)kWh-1
|
Total Daily Average Flow,
m3day-1
|
Number of Facilities
|
Power,
kWhYr-1
|
Annual CO2(e), TonnesYr-1
|
Estimated Annual CO2(e) For LEDs in 2025, TonnesYr-1
|
BC
|
7.3
|
129784
|
45
|
550521
|
4.02
|
3.03
|
AB
|
590
|
975988
|
33
|
4139965
|
2443
|
1842
|
SK
|
580
|
186428
|
6
|
790794
|
459
|
346
|
MB
|
1.1
|
199519
|
12
|
846324
|
0.931
|
0.702
|
ON
|
25
|
1250231
|
172
|
5303253
|
133
|
100
|
QC
|
1.5
|
1084258
|
39
|
4599226
|
6.90
|
5.20
|
NB
|
290
|
64587.8
|
28
|
273969.8
|
79.5
|
59.9
|
NL
|
24
|
16225.2
|
8
|
68824.37
|
1.65
|
1.25
|
NS
|
670
|
280726.5
|
59
|
1186549
|
795
|
599
|
PE
|
0
|
36575
|
19
|
155144.5
|
0
|
0
|