Energy Profiles of Nine Water Treatment Plants in the Salt Lake City Area of Utah and Implications for Planning, Design, and Operation

Sowby, Robert B.; Thompson, Matthew J.

doi:10.1061/(ASCE)EE.1943-7870.0001888

Free access

Technical Papers

May 13, 2021

Energy Profiles of Nine Water Treatment Plants in the Salt Lake City Area of Utah and Implications for Planning, Design, and Operation

Authors: Robert B. Sowby, M.ASCE [email protected], and Matthew J. Thompson https://orcid.org/0000-0001-5023-0179 [email protected]Author Affiliations

Publication: Journal of Environmental Engineering

Volume 147, Issue 7

https://doi.org/10.1061/(ASCE)EE.1943-7870.0001888

PDF

Abstract

This study examined primary data on the energy use (electricity and natural gas) of nine surface water treatment plants in and around Salt Lake City, Utah, and how their energy use relates to their sizes, processes, and climates. The energy intensities per unit of treated water, averaged over a 3-year period, ranged from 0.059 to

0.565 kW \cdot h / m^{3}

[

220 - 2,140 kW \cdot h / million gal.

(MG)] with a weighted average of

0.162 kW \cdot h / m^{3}

(

610 kW \cdot h / MG

), concurring with the literature. The data confirmed economies of scale: larger plants and higher production had lower energy intensities. The analysis found energy differences for certain disinfection processes; ozone and ultraviolet disinfection required more energy than did chlorination. Significant energy baseloads (presumably for ventilation, lighting, and other building systems) were observed, even when little or no water was being treated. All sites had significant natural gas use, which was attributed mostly to space heating—33% to 67% of total energy—which should not be overlooked in energy management programs. Additional recommendations for sustainable planning, design, and operation of water treatment plants were discussed.

Introduction

Energy is one of the largest operating costs for public water supply, a cost that is increasing due to scarcer water supplies (stressed by both population growth and climate change), rising prices for electricity, and stricter water treatment standards (USEPA 2008). Therefore, energy management is a subject of increasing importance for water suppliers. Much guidance has emerged in recent years to address energy use as part of broader water sustainability efforts (EPRI 2002, 2009, 2013; USEPA 2008; NYSERDA 2010; Liu et al. 2012; AWWA 2016; Badruzzaman et al. 2020).

Many researchers have studied the energy requirements of water supply at various scales. These include studies of foreign countries (Wakeel et al. 2016), the US (Twomey and Webber 2011; Plappally and Lienhard 2012; Sanders and Webber 2012; USDOE 2014), states and regions (Wilkinson 2000; Wilkinson et al. 2006; Bennett et al. 2010a, b; Blanco et al. 2012), individual water utilities (Elliott et al. 2003; ISAWWA 2012; Young 2015; Sowby and Burian 2017a, b, 2018, 2020; Chini and Stillwell 2018; Sowby 2018b), and components such as treatment, distribution, and end-use (EPRI 2002, 2009, 2013; Siddiqi and Fletcher 2015). Other studies have noted the general difficulty of acquiring adequate local water and energy data for these purposes (Chini and Stillwell 2017; Sowby et al. 2019a; Badruzzaman et al. 2020).

The energy use of water treatment plants (WTPs) specifically has been studied often, including a notable review by Loubet et al. (2014) and contributions by others (Racoviceanu et al. 2007; Plappally and Lienhard 2012; Lam et al. 2017; Molinos-Senante and Sala-Garrido 2017, 2018; Sala-Garrido and Molinos-Senante 2020). Many of these works reported a normalized energy intensity for water treatment in kilowatt-hours per cubic meter of water treated (

kW \cdot h / m^{3}

). The average energy intensities from these studies ranged from 0.03 to

0.37 kW \cdot h / m^{3}

[

110 - 1,400 kW \cdot h / million gal.

(MG)] globally. Such a metric is simple to compute but admittedly limited, because it does not fully describe the WTP’s geographic setting, processes, source water quality, treatment objectives, age, size, seasonality, or many other factors that might influence its energy use. The literature acknowledges economies of scale, in which larger WTPs use less energy per unit of water treated than smaller WTPs, because incremental flows do not change the baseload of building, ventilation, lighting, and other nonprocess systems. A water quality factor also is apparent: poorer source-water quality and/or higher treatment standards require more energy.

Despite considerable research to date, very few empirical studies of operating WTPs have been completed. Using primary data, this research examined the energy use and energy intensity of nine conventional surface water treatment plants owned by various municipalities and water districts in and around Salt Lake City, Utah to better understand key drivers impacting their variability. The empirical energy intensities from the sample were compared with values from the literature, and a regression model was developed to profile the sites according to water production, plant size, treatment processes, and climate. Implications for planning, design, and operation were discussed.

Methods

Study Area and Selected Sites

The greater Salt Lake City area faces challenges that are prompting decisions about sustainable water infrastructure in general and about replacing aged WTPs in particular. First, Utah’s population is expected to almost double from 3.0 million in 2015 to 5.8 million in 2065, with 600,000 of those new residents expected in the Salt Lake City area (Kem C. Gardner Policy Institute 2017). The dramatic increase in population will stress existing water supplies and infrastructure. Second, the projected effects of climate change in Utah by 2050 relative to present conditions include a temperature increase of 1.3°C, reductions in mountain snowpack, and peak runoff occurring 1 month earlier, all of which will affect the availability and timing of surface water supplies and the placement and design of WTPs (Barnett et al. 2005; Gillies et al. 2012; USEPA 2015; Khatri et al. 2018; USGCRP 2018). Third, the average price of industrial electricity in Utah has increased by 70% since 2001 (EIA 2020), motivating water utilities to find ways to save energy in an era in which the public expects efficient use of energy, money, and other resources. Finally, as part of an estimated $18 billion needed investment to repair and replace aging water infrastructure statewide (Prepare60 2013), some local WTPs are over 60 years old and quickly are approaching the end of their useful life. Their imminent replacement offers an opportunity to consider modern energy efficiency in the new designs.

Accordingly, this geographic area is a logical place to focus the study of WTPs, their energy profiles, and their local context to inform upcoming decisions. The findings also may guide similar decisions elsewhere. To enable comparability, nine WTPs were selected according to the following criteria:

•

the plant is located along the Wasatch Front, a populous strip of land at the base of the Wasatch Mountains in the Salt Lake City area of northern Utah;

•

the plant treats gravity-fed surface water;

•

the plant produces potable water per the USEPA’s drinking water standards;

•

the plant has at least 3 years of consecutive monthly records of energy use (electricity and gas) and water production;

•

the plant has not undergone a major upgrade or construction project during the period that would alter the energy use;

•

the plant’s electric load is primarily attributable to the treatment operations, or the plant’s other electric loads (like finished water pumping) can be separated; and

•

the plant’s processes can be adequately described with a few options.

Fig. 1 shows the site locations. They are identified as Site A, Site B, and so forth, to protect sensitive infrastructure data.

Fig. 1. Water treatment plant sites and study area.

Process Characterization

The characterization of water treatment processes followed Table 4–2 of EPRI (2013), with a few modifications. To facilitate responses, irrelevant characteristics were omitted (such as those regarding reverse osmosis). Clarifying characteristics were added for plant capacity, water source, finished water pumping, and nonplant energy use. The result was a survey spreadsheet, primarily with yes/no questions, completed by the water suppliers (Fig. S1).

Energy and Water Data

Energy use (electricity and natural gas) and water production were provided by the water suppliers. The data constituted monthly records over 3 calendar years. The 3 consecutive years varied among the plants, but all were between 2013 and 2019. Natural gas use in dekatherms was converted to equivalent kilowatt-hours (

1 dth = 293 kW \cdot h

) and was combined with electricity use (also in kilowatt-hours) for total energy use. When possible, water production values were compared with public records from the Utah Division of Water Rights (UDWR 2020) to ensure accuracy. When a survey response indicated significant finished water pumping, additional information was collected to separate its energy consumption from that of the treatment processes.

Source Water Quality

Raw water quality is similar across all nine sites. All receive source water from the same regional watershed, all have strict source protection practices, and all are the first users of such water, so there is little contamination from upstream uses. The water originates as snowpack in the Wasatch Range and is routed through mountain reservoirs, where solids settle out. In the process of providing other data, plant operators indicated that their raw water quality is very consistent throughout the year. Aside from occasional changes in chemical dosing (and consequent sludge production), incoming water quality parameters are not likely to significantly influence energy use, especially compared with variables such as plant size and flow rate. Accordingly, source water quality would not have helped to differentiate energy use at these particular sites and was not analyzed in this study. However, it could be important elsewhere, and should be considered.

Climate Data

Given Utah’s diverse weather conditions throughout the year—ranging from hot, arid summers to cold, wet winters—it was important to examine their correlation with plant energy use (e.g., heating) and water demand. Average monthly air temperature and monthly total precipitation over the respective historical periods were extracted for each site from gridded data (PRISM Climate Group 2020a, b). Heating degree days (HDDs) and cooling degree days (CDDs) were computed from the monthly gridded data relative to a base of 18°C (65°F). HDD and CDD data from National Oceanic and Atmospheric Administration (NOAA) and National Weather Service (NWS) weather stations were considered, but those data are incomplete and do not adequately represent the heating and cooling at these sites, many of which are at much higher elevations than the weather stations.

Regression Model

Although energy intensity alone is a useful metric, this study assessed the effects of facility, process, and climate variables on a WTP’s energy use. Ordinary least squares (OLS) linear regression models of monthly energy use and monthly energy intensity were developed for this purpose, similar to Carlson and Walburger (2007) and Sowby and Burian (2018) but specific to these nine WTPs. The variables in Table 1 were considered as possible explanatory variables. Model specification was iterative and stepwise. A new variable was retained only if its

p

-value was less than 0.05 (i.e., there was less than 5% chance that the variable’s correlation with energy use was random). The approach also sought to maximize the adjusted

R^{2}

value and minimize the RMS error (RMSE).

Table 1. Water treatment plant process and climate data

Site^a	A	B	C	D	E	F	G	H	I
Years of record	2017–2019	2017–2019	2016–2018	2016–2018	2016–2018	2013–2015	2016–2018	2016–2018	2016–2018
Capacity ( $m^{3} / day$ )	568,000	265,000	681,000	76,000	379,000	174,000	170,000	57,000	144,000
Capacity [ $million gal. / day$ (MGD)]	150	70	180	20	100	46	45	15	38
Ballasted sand	—	—	—	X	—	—	—	—	—
Powdered activated carbon	—	—	X	X	—	X	—	—	—
Repumping within plant	X	X	X	—	X	X	—	—	—
Backwash pumps	X	X	X	—	—	X	X		X
Residuals pumping	—	X	—	—	X	—	—	—	—
Thickened solids pumping	—	—	X	—	X	X	—	—	—
Hypochlorite disinfection	—	X	—	X	—	X	—	X	—
Chlorine gas disinfection	—	—	X	—	X	—	X	—	X
Ozone disinfection	X	X	—	—	X	X	—	—	—
Ultraviolet disinfection	—	X	—	—	—	X	—	—	—
Heating degree days^b,^c	2,960	2,930	2,910	2,790	3,080	3,160	3,280	3,680	3,080
Cooling degree days^b,^c	434	452	527	575	425	439	236	150	395
Average air temperature^b (°C)	11.4	11.5	11.7	12.2	11.0	10.8	9.9	8.6	10.9
Precipitation^c ( $m / year$ )	0.62	0.51	0.38	0.50	0.51	0.55	0.64	0.64	0.57

a

All sites have coagulation, flocculation, and sedimentation processes.

b

Average year, based on 3 years of record.

c

Average year, base 18°C.

Multicollinearity between predictor variables was assessed with variance inflation factors (VIFs), and models with exceedingly high levels (

> 10

) were omitted (Alin 2010). The models’ ability to predict energy use and energy intensity was compared using the second-order Akaike information criterion (AIC) (Elnakar and Buchanan 2020). Lastly, a sensitivity analysis was performed to evaluate magnitude of influence and variability of each predictor variable. As with any statistical model, the method required an artful balance of plausibility and significance. The models were used as exploratory tools for identifying significant variables influencing energy use in general, but also could be used for future benchmarking of the nine local WTPs (Table S1).

Given the wide range of values for energy use and water production at these sites, logarithmic transformation of both parameters was necessary to linearize the data, i.e., the natural logarithms of each were substituted for the original values, as in Carlson and Walburger (2007) and Sowby and Burian (2018). Monthly energy use and energy intensity were evaluated as dependent variables, as conducted similarly by Hanna et al. (2018) for small water resource recovery facilities. The independent variables investigated were a blend of flow parameters, process characteristics, and climate-related factors. Including all variables initially could lead to overfitting because the total sample size was limited and contained many binary variables. To address this issue, variable selection for Models 1 and 3 included only the disinfection processes, because they are known to be a large energy use operation in drinking water plants. Models 2 and 4 then excluded disinfection processes, and instead included additional pumping and other processes with the facility.

Results and Discussion

Water Treatment Plant Characteristics

Table 1 summarizes the nine WTPs’ characteristics. All have coagulation, flocculation, and sedimentation processes; the table identifies additional processes as well as each plant’s capacity and climate setting (HDD, CDD, air temperature, and precipitation).

Water Treatment Plant Energy Use

Table 2 summarizes the annual electricity use, natural gas use, water production, and energy intensity data. The observed average energy intensity values of

0.059 - 0.565 kW \cdot h / m^{3}

(

220 - 2,140 kW \cdot h / MG

) were consistent with the ranges given in the literature cited previously. Interestingly, there was almost an order of magnitude difference even across only nine local sites. Natural gas use was a significant component for each facility, ranging from 33% (Site I) to 67% (Site G) of the total energy use.

Table 2. Water treatment plant energy use and energy intensity

Site	Average annual electricity use^a ( $kW \cdot h$ )	Average annual natural gas use^a ( $kW \cdot h$ )	Average annual water production^a ( $m^{3}$ )	Average annual energy intensity^a ( $kW \cdot h / m^{3}$ )
A	6,781,000	10,590,000	80,466,000	0.216
B	3,200,000	3,422,000	11,717,000	0.565
C	2,901,000	2,969,000	80,981,000	0.072
D	808,000	876,000	11,060,000	0.152
E	3,009,000	2,034,000	36,386,000	0.139
F	3,043,000	1,734,000	23,506,000	0.203
G	813,000	1,615,000	8,033,000	0.302
H	354,000	443,000	6,246,000	0.128
I	1,051,000	511,000	26,494,000	0.059

a

Averaged over 3 years.

Fig. 2 plots the plants’ energy use and energy intensity versus their water production. In general, energy use increased with water production, but energy intensity decreased with water production. The regression model described subsequently further explored the significance of these factors and economies of scale.

Fig. 2. (a) Water treatment plant energy use and water production by site; and (b) water treatment plant energy intensity and water production by site.

The observed energy intensities are not energy efficiency benchmarks and should not be used to compare performance among the nine plants. They are merely empirical statements of the energy requirements, with no comparison to any performance standard. Indeed, as mentioned previously, many variables contribute to a WTP’s energy profile, and judging the performance requires a more sophisticated approach.

Fig. 3 shows the monthly energy use by source, water production, and HDD for each of the nine sites (beginning in January). With minor exceptions, all sites had clear relationships between increased natural gas and HDD due to space heating. Electricity use increased with water production in almost all cases. For most facilities there is a significant baseload of electricity use regardless of water production, most notably at Sites B, C, and D. This likely is attributable to auxiliary loads such as lighting, air conditioning, computers, and mechanical equipment whose energy use does not depend on the flow. Further assessment could delineate the specific unit process energy use and areas of greatest opportunity for improvement.

Fig. 3. Water treatment plant energy use by type, water production, and HDD over time.

Effects of Size, Processes, and Climate

Several of the regression models revealed key factors influencing energy use and the energy intensity of drinking water treatment systems. Table 3 lists the energy use and energy intensity model coefficients of the various models.

Table 3. Monthly energy and energy intensity regression models

Variable	Energy use^a		Energy intensity^a
Variable	Model 1 coefficient	Model 2 coefficient	Model 3 coefficient	Model 4 coefficient
Intercept	4.88	2.58	4.88	7.56
Continuous variables
Flow
Water production volume^a (m3)	0.52	0.66	$- 0.48$	$- 0.77$
Design capacity (%)	$- 1.19$	$- 1.53$	$- 1.19$	$- 0.59$
Climate and building
Approximate floor area ( $m^{2}$ )	$3.82 \times 10^{- 5}$	—	$3.82 \times 10^{- 5}$	$9.87 \times 10^{- 5}$
Heating degree days (base 18°C)	$1.62 \times 10^{- 3}$	$1.74 \times 10^{- 3}$	$1.62 \times 10^{- 3}$	$1.55 \times 10^{- 3}$
Binary variables (1 = present, 0 = absent)
Disinfection type
Sodium hypochlorite	$- 0.63$	—	$- 0.63$	—
Gaseous chlorine	$- 0.54$	—	$- 0.54$	—
Ozone	0.18	—	0.18	—
Ultraviolet	0.41	—	0.41	—
Other processes
Repumping within plant	—	0.90	—	0.41
Ballasted sand	—	0.67	—	0.64
Powdered activated carbon	—	$- 0.21$	—	—
Regression statistics
Adjusted $R^{2}$	0.85	0.82	0.83	0.86
RMS error (RMSE)	0.39	0.43	0.39	0.35
Akaike information criterion (AIC)	285	218	285	218

a

Variable transformed with natural logarithm.

In every model, an economy of scale appeared, with the volume of water production leading to larger energy use but lower energy intensity (Fig. 2). Furthermore, plants operating at or near their design capacity had both lower energy use and lower energy intensity relative to those operating far below their design capacity. These observations are consistent with the literature (Molinos-Senante and Sala-Garrido 2017, 2018; Sala-Garrido and Molinos-Senante 2020). The type of disinfection process employed also correlated with energy use and energy intensity, with ozone and ultraviolet disinfection contributing to higher energy use relative to sodium hypochlorite and gaseous chlorine, consistent with the literature (WateReuse Research Foundation 2012). In Models 2 and 4, repumping of water within the facility also led to increased energy use and energy intensity.

The negative correlation between hypochlorite/gaseous chlorine and energy use or energy intensity is because these facilities on average had lower energy use relative to ozone and ultraviolet disinfection. As discussed subsequently, the sensitivity analysis of this variable indicated that its influence on the overall estimated energy use and energy intensity was relatively minimal compared with other factors.

Given the large quantity of natural-gas use, building- and climate-related variables were significant in all models investigated. Both building floor area and HDD appeared to be good predictors of a facility’s overall energy use and energy intensity. Both variables have been observed to be significant in unit process energy assessments (Thompson et al. 2020) and energy benchmarking studies of water resource recovery facilities (Hanna et al. 2018). HDD also correlated with the energy profile, in which colder climates contribute to greater heat loss and consequentially higher energy use. Although the plants are located in the same general geographic area, their specific locations and elevations differ enough to affect the HDD and therefore the energy use.

Fig. S2 shows the model fit for the four regression models. The sensitivity analysis showed that the water production was overwhelmingly the largest variable impacting the predicted energy use or energy intensity, whereas all other factors only attributed minor influences. Percentage design capacity, HDD, and climate-controlled floor area all were more influential than the process-specific components. The model comparison using AIC showed that Models 2 and 4 only slightly outperformed Models 1 and 3 in predicting energy use and energy intensity.

Recommendations for Planning, Design, and Operation

The foregoing analysis suggests the following recommendations for planning, design, and operation to improve energy performance:

•

Scale up. Due to the observed economy of scale, larger, centralized water treatment plants require less energy per unit of water treated. However, costs of additional pumping, consolidation, and piping also must be considered.

•

Avoid oversizing facilities. Plants operating near their design capacity are less energy intensive due to economy of scale, because some operations have a more constant energy use irrespective of flow rate (e.g., space heating, lighting, controls). Avoid constructing large facilities the capacity of which will not be used fully until long into the future. Rather, consider a modular design in which treatment trains can be added as needed over time. Furthermore, reducing building size and volume will reduce the energy needed to heat the space.

•

Insulate. Heating and cooling can consume significant energy, often more than treatment. Improving building insulation—particularly at roll-up doors, windows, and hatches—at existing and new facilities is an investment that will lower the operational energy use. Likewise, where freezing pipes are a concern, consider insulating the pipes rather than the room.

•

Lower heating setpoints. Many areas in a water treatment plant are unoccupied and are visited only briefly during operator rounds. A heating setpoint of about 10°C (50°F) is adequate for such spaces. Programmable thermostats are recommended. Rooms with chemical systems may need higher setpoints to avoid undesirable changes in chemical properties such as vapor pressure.

•

Submeter. Submetering every motor is not practical or necessary, but the ability to separate and analyze major energy uses, especially nonprocess loads such as finished water pumping, is critical to energy management.

•

Minimize auxiliary loads. The observed energy baseloads, even when a WTP treated no water, suggest opportunities to minimize energy use of lights, HVAC, and other building systems, perhaps with occupancy sensors or timers.

•

Optimize motor speeds. The goal of every WTP is to reliably produce high-quality water. Often there are opportunities to decrease mixer and pump speeds (or install variable-frequency drives to do so) to meet the same water quality objectives with less energy (running low and slow).

•

Consider disinfection energy. Not all disinfection methods are equal in terms of energy use. Ozone and ultraviolet disinfection require more energy than chlorination, but also have benefits that chlorination does not. Energy is admittedly only one of many considerations when selecting disinfection methods; engineers will choose the one that best meets the water quality goals.

•

Develop and track key performance indicators (KPIs). Developing energy related KPIs for the overall facility and energy-intensive processes can help WTPs track both the impact of improvements and reductions in efficiency over time. Concurrently, the regression models developed in this study can help track how the facility is performing on average relative to other sites. An example calculation is provided in the Supplemental Materials.

Conclusions

This study examined the monthly energy use of nine water treatment plants around Salt Lake City over a 3-year period. The energy intensities ranged from 0.059 to

0.565 kW \cdot h / m^{3}

(

220 - 2,140 kW \cdot h / MG

) of finished water, and reflected the variety of WTP sizes, production, processes, and climate conditions in the sample. In general, larger plants require less energy per unit of finished water, as do plants that operate near their design capacity. The range of energy intensities and economy of scale both confirm the findings of previous research.

One notable finding of this study is that all plants had significant electricity and/or natural gas baseloads even when producing little or no water. Lighting, HVAC, and other building systems therefore deserve further attention at WTPs, but such discussion is underrepresented in the literature. Space heating accounted for 33%–67% of the total energy use in the sample, suggesting that opportunities to improve insulation and heating efficiency should not be overlooked in energy management programs. The findings could complement efforts to improve operational energy efficiency, such as those described by others (Jones and Sowby 2014; Sowby et al. 2017; Sowby 2018a; Sowby et al. 2019b; Thompson et al. 2020).

Recommendations for managing energy use include regionalizing water treatment, avoiding overdesign, better insulating water facilities, controlling heating of unoccupied spaces, submetering major electric loads, minimizing auxiliary loads, optimizing motor speeds, considering disinfection energy, and tracking performance.

Supplemental Materials

File (supplemental_materials_ee.1943-7870.0001888_sowby.pdf)

Download
587.35 KB

Data Availability Statement

Some or all data, models, or code that support the findings of this study are available from the corresponding author upon reasonable request: Table of anonymized water treatment plant water production, energy use, process characterization, and climate variables.

Acknowledgments

The authors thank the following organizations and employees for contributing data to this study: Central Utah Water Conservancy District (Russ Franklin and Rick Maloy), Jordan Valley Water Conservancy District (Steve Blake, Todd Schultz, and Ray Stokes), Metropolitan Water District of Salt Lake and Sandy (Nathan Scown and Matt Tietje), Salt Lake City Department of Public Utilities (Marian Rice and Jesse Stewart), and Weber Basin Water Conservancy District (Jon Parry and Brody Tait). The authors also thank three anonymous reviewers for their helpful suggestions.

References

Alin, A. 2010. “Multicollinearity.” Wiley Interdiscip. Rev. Comput. Stat. 2 (3): 370–374. https://doi.org/10.1002/wics.84.

Abstract

Introduction

Methods

Study Area and Selected Sites

Process Characterization

Energy and Water Data

Source Water Quality

Climate Data

Regression Model

Results and Discussion

Water Treatment Plant Characteristics

Water Treatment Plant Energy Use

Effects of Size, Processes, and Climate

Recommendations for Planning, Design, and Operation

Conclusions

Supplemental Materials

Data Availability Statement

Acknowledgments

References

Information

Published In

Copyright

History

Permissions

Authors

Affiliations

Metrics

Citations

Download citation

Cited by

Figures

Other

Share

Copy the content Link

Share with email

Share

Request Username

Create a new account

Change Password

Password Changed Successfully

Verify Phone

Congrats!