Expansion of the Petroleum Refinery Life Cycle Inventory Model to

Article pubs.acs.org/est

Cite This: Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Expansion of the Petroleum Refinery Life Cycle Inventory Model to Support Characterization of a Full Suite of Commonly Tracked Impact Potentials Ben Young,† Troy Hottle,† Troy Hawkins,‡ Matthew Jamieson,§ Gregory Cooney,*,§ Kavan Motazedi,¶ and Joule Bergerson¶ †

Franklin Associates, A Division of Eastern Research Group, Lexington, Massachusetts 02421-3136, United States Energy Systems Division, Argonne National Laboratory, Lemont, Illinois 60439-4801, United States § National Energy Technology Laboratory, Pittsburgh, Pennsylvania 15236-0940, United States ¶ University of Calgary, Calgary, Alberta T2N 1N4, Canada

Downloaded via UNIV OF MASSACHUSETTS AMHERST on February 5, 2019 at 07:38:48 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

‡

S Supporting Information *

ABSTRACT: This study updates the Petroleum Refinery Life Cycle Inventory Model (PRELIM) to provide a more complete gate-to-gate life cycle inventory and to allow for the calculation of a full suite of impact potentials commonly used in life cycle assessment (LCA) studies. Prior to this update, PRELIM provided results for energy use and greenhouse gas emissions from petroleum refineries with a level of detail suitable for most LCA studies in support of policy decisions. We updated the model to add criteria air pollutants, hazardous air pollutants, releases to water, releases to land, and managed wastes reflecting 2014 reported releases and waste management practices using data from the U.S. Environmental Protection Agency Greenhouse Gas Reporting Program, National Emissions Inventory, Discharge Monitoring Reports, and Toxic Release Inventory together with process unit capacities and fuel consumption data from the U.S. Energy Information Administration (U.S. EIA). The variability of refinery subprocess release factors is characterized using log-normal distributions with parameters set based on the distribution of release factors across facilities. The U.S. EPA Tool for the Reduction and Assessment of Chemical and Environmental Impacts life cycle impact assessment (LCIA) method is used together with the updated inventory data to provide impact potentials in the PRELIM dashboard interface. Release inventories at the subprocess level enable greater responsiveness to variable selection within PRELIM, such as refinery configuration, and allocation to specific refinery products. The updated version also provides a template to allow users to import PRELIM inventory results into the openLCA software tool as unit process data sets. Here we document and validate the model updates. Impact potentials from the national crude mix in 2014 are compared to impacts from the 2005 mix to demonstrate the impact of assay and configuration on the refining sector over time. The expanded version of PRELIM offers users a reliable, transparent, and streamlined tool for estimating the effect of changes in petroleum refineries on LCIA results in the context of policy analysis.

■

INTRODUCTION

coming years. Domestic crude oil extraction has grown 75% in the past decade from 5.1 million bpd in 2006 to 8.9 million bpd in 2016.9 The expansion of oil extraction from shale formations is changing the types of crude being processed by U.S. refineries.10−12 The U.S. Energy Information Administration (U.S. EIA) Annual Energy Outlook 2018 projects continued growth in light tight oil primarily from west Texas and eastern New Mexico.13 Changes in domestic petroleum

Petroleum refining is a major global industry with significant environmental impacts. In 2014, U.S. refineries generated approximately 175 MMT CO2 eq, or approximately 3% of U.S. total greenhouse gases (GHG).1,2 In addition to GHGs, refineries generate other criteria air pollutants (CAPs),3,4 hazardous air pollutants (HAPs),5 wastewater effluent,6 and solid wastes,7 which contribute to the environmental burden of refining activities.2,8 Refinery operations have seen significant changes in recent years because of changing crude assays (the chemical composition of the crude oil), fuel specifications, and environmental requirements, and these changes are expected to continue in © XXXX American Chemical Society

Received: October 3, 2018 Revised: December 28, 2018 Accepted: January 7, 2019

A

DOI: 10.1021/acs.est.8b05572 Environ. Sci. Technol. XXXX, XXX, XXX−XXX

Article

Environmental Science & Technology

features enable users to evaluate the environmental implications of business and policy decisions at the refinery. For example, changes in drilling and recovery methods will affect crude oil characteristics and thus processing intensity and emissions at the refinery. Policies such as those requiring the use of ultralow sulfur diesel place new demands on refineries. These business and policy decisions demonstrate the need to be able to evaluate potential trade-offs between impacts on a life-cycle basis. As an update to the 2014 GHG petroleum baseline developed by Cooney and colleagues,11 we compare in this analysis gate-to-gate refinery impacts from the national crude mix in 2014 to the mix from 2005. To our knowledge, PRELIM v1.3 is the first instance of using facility-level data to create an open-source, configurable subprocess environmental inventory model for an entire sector. The inventories created using PRELIM v1.3 are intended for use in conjunction with established LCIA methods such as the Tool for Reduction and Assessment of Chemicals and Other Environmental Impacts (TRACI)21 or ReCiPe.22 This paper serves as a first step toward assessing the environmental outcomes associated with changing feedstocks, environmental constraints, and markets affecting the petroleum refining and transportation sectors. Here we describe the way we addressed data availability, harmonization of data sets, the assignment of inputs and outputs to the subprocesses included in PRELIM, and the integration of the updated inventories with the PRELIM model.

extraction translate to different crude oil compositions entering U.S. refineries and therefore release inventories (the aggregated accounting of chemical flows to the environment).14 Additionally, the refinery product slate (the type and quality of refined products produced) is changing, for example in response to increased demand for ultralow sulfur diesel following the tightening of sulfur standards of highway and marine fuels (40 CFR Part 80). Together, these changes highlight the need for modeling that can reflect variability in crude oil composition and processing intensity at refineries. In updating the U.S. petroleum life cycle GHG baseline, Cooney and colleagues11 document various life cycle assessments (LCAs) in the literature evaluating GHG emissions associated with petroleum products and the limitations of those studies due to data availability and transparency, geographical relevance, and comparisons that do not represent the diversity in refinery configurations or crude oil assays. Their findings point to a shift in the GHG emissions from petroleum refineries due to the changing mix of crude oils consumed in the U.S. Multiple other LCA studies of transportation fuels call for more process granularity at the refinery to better reflect allocation across coproducts. Wang and colleagues stress the benefits of allocating energy demands to refinery subprocesses as it better reflects differences in processing intensity of refinery coproducts and meets the recommendation of the International Standard Organization Standard for Life Cycle Assessment (ISO 14044) to avoid allocation where possible by distinguishing releases from subprocesses and assigning those subprocesses to specific (intermediate) products.15 Michalek and colleagues note the need for refinery models which can effectively reflect production trade-offs necessary to support more comprehensive life cycle assessments of transportation fuels.16 Finally, characterizing refinery releases at the subprocess level allows for models predicting the effects of changes to refinery operations, for example due to shifts in crude composition.17 The Petroleum Refinery Life Cycle Inventory Model (PRELIM) is an established, freely available, open-source tool for estimating energy use by and GHG emissions from petroleum refining.18 Model parameters allow for scenario analyses based on 11 refinery configurations. Gate-to-gate results for refinery products are assigned based on subprocess flows and allocation of subprocess coproducts on an energy, mass, hydrogen content, or economic basis (Figure S1). The PRELIM database contains 144 distinct crude oil assays that account for differences in the crude distillation curve, API gravity, sulfur content, hydrogen content, and carbon residue.19 PRELIM allows users to explore the sources of GHG emissions within refineries, which are predominantly associated with energy demand such as heat, electricity, hydrogen, and catalyst coke burnoff within the process units.20 Here we provide an updated version of PRELIM (v1.3) with an expanded inventory of environmental releases of elementary flows, including GHGs, CAPs, HAPs, releases to water, releases to soil, and managed wastes from within the refinery gate. The expanded inventory is based on publicly available data and is responsive to refinery configuration and assay selection. This expanded inventory enables more complete life cycle impact assessments (LCIAs) for refinery coproducts that reflect differences in process throughput, crude assay, and refinery configuration to more accurately predict changes in life cycle impacts from the refinery sector over time. These

■

METHODS Expanded environmental release inventories were added to the existing process unit modeling in PRELIM reflective of the 2014 U.S. national average releases for each unit. Environmental releases from refineries in 2014 were sourced from the National Emissions Inventory (NEI), Greenhouse Gas Reporting Program (GHGRP), Toxic Release Inventory (TRI), and Discharge Monitoring Reporting program (DMR).23−26 Activity data for process units at each refinery were estimated based on operating capacities from U.S. EIA’s annual Refinery Capacity Report (RefCap)27 and capacity utilization rates by Petroleum Administration for Defense Districts (PADD) subregion.28 In the case of fuel combustion for heat-related subprocesses, activity levels were estimated by PADD based on fuel use.29 The relationships between these data sets and the model’s elementary flows, as well as the number of elementary flows of each type reported at refineries, are provided in Table 1 with additional detail in the Supporting Information. Environmental Releases. Data reported by refineries were compiled and harmonized to generate an expanded inventory of inputs and outputs for each facility. The approach to compiling release data is similar to the one described by Cashman and colleagues30 and builds on earlier work to characterize releases from refineries at the subprocess level.2,31 Subprocess designations reflect the source of the environmental release, such as heat-related combustion releases from heaters or boilers; process unit releases from the main line of petroleum product process units; or other sources such as flares, evaporative emissions from tanks, and other fugitive emissions. Correspondences were created to match facilities, elementary flows, environmental compartments, and refinery subprocesses across data sets (Table S1). Where a flow is reported by a refinery in more than one data set, the maximum reported release is assumed. B


Article


Table 1. Data Sources Used to Characterize Refinery Elementary and Technosphere Flows in the Data Mining Approacha

a

Level of detail provided is shown below each dataset: Subprocess data for each facility (Sub), facility totals only (Facility), or PADD totals. Values represent the total number of substances used from each dataset, and the values in parentheses represent the estimated percent of U.S. petroleum throughput reflected by the flows used from each dataset. Where flows are sourced from more than one dataset, data sources are ordered by decreasing preference from left to right.

Under the Greenhouse Gas Reporting Rule (40 CFR Part 98), regulated facilities report releases of GHGs to U.S. EPA, which are published annually in the GHGRP. Relevant releases are reported by refineries under Subpart Y, Petroleum Refineries; Subpart P, Hydrogen Production; and Subpart C, General Stationary Fuel Combustion Sources. We assigned releases to subprocesses based on the subpart to which they were reported and using descriptions reported by refinery operators. The emissions reported by refineries to the GHGRP are limited to carbon dioxide, methane, and nitrous oxide, so there was no overlap in flows reported in the GHGRP and the other GHGs, such as the hydrofluorocarbons and chlorofluorocarbons reported in the NEI. Facilities report releases of CAPs and HAPs in the NEI at individual emissions sources by eight-digit source classification codes (SCCs). These SCCs identify the emissions source and, for combustion processes, the fuel type. We assigned releases to subprocesses based on SCCs (see the Supporting Information). Releases of HAPs are also reported in the TRI (Table 1). Because of the subprocess specificity in the NEI, it is the preferred source when flows overlap (Table S3 and Figure S2).30,32 Releases to water of refinery effluent containing metals, hydrocarbons, and organic wastes are reported to the TRI and DMR. Releases are reported as facility totals but were assigned to refinery process units according to expected wastewater generation rates (Table S6).33 We assigned certain flows such as metals and dioxins to specific process units,33 such as desalting and the water wash following catalyst regeneration, from which they are known to originate (Table S5). In the case where refineries report liquid waste transferred for treatment at an off-site treatment facility, we assume a treatment efficiency of 80% resulting in 20% of toxins sent for off-site wastewater treatment being released to water.6,34 Refinery operators report the amounts of toxic wastes, such as spent catalysts managed on-site or sent off-site for further management, to the TRI (Table S7). We estimated the releases to air, water, and soil associated with these managed waste flows based on the management practices specified by the TRI.35 Each management practice was assigned a release fraction to an environmental compartment based on the most likely pathway of release. The release fractions were based on literature values for expected releases from each step involved in each management practice (Table S8).7,36,37 For example, releases from incineration of toxic wastes off-site assume a 1% loss to air due to releases associated with handling,

Table 2. Activity Data and Facility Count for Release Factor Development data set facility count (M) release factor (F)

activity data (A)

GHGRP

NEI

heat-related air release factors process unit air release factors fluid catalytic cracker (FCC) coker catalytic reformer sulfur plant steam methane reformer asphalt plant process unit water release factors support services air release factors miscellaneous air release factors releases from managed wastes

refinery gas and natural gas consumption (MJ) process unit throughput

127

122

FCC feed (m3)

93

77

coker feed (m3) reformer feed (m3) sulfur (kg)

54 98

33 25

93

71

hydrogen (m3)

50

9

asphalt and road oil (m3) process unit throughput

8

−

refinery throughput (m3 crude) refinery throughput (m3 crude) refinery throughput (m3 crude)

122

TRI

DMR

106

86

122 117 117

transportation, and efficiency of combustion.38 For off-site metals recycling, the efficiency associated with metals recovery from catalysts was used to approximate a 3% disposal rate to landfill. From this subset, a 1% loss over the life of the landfill results in a final release of 0.3% to water via landfill leachate.7 Activity Data. We created release factors for each environmental release based on appropriate activity levels. Throughput or production is used for process unit releases; fuel consumption is used for heat-related releases, and refinery crude throughput is used for support services, miscellaneous, and managed waste. We estimated release factors (F) for each elementary flow (i) from subprocess (s) using two separate approaches to better reflect uncertainty surrounding unreported flows. Under the first approach (F′), total emissions (M) of an elementary flow from a subprocess were divided by total activity (A) of all facilities (j) that report that specific flow from that subprocess C


Article


Figure 1. Contribution analysis by process type to photochemical ozone formation potential by refinery. Impacts are normalized to utilized operating capacity and ranked from highest to lowest. Excluded facilities and flows are removed. Additional impact categories are included in the Supporting Information.

each process type are provided in Figure S9. This approach is informed by Sengupta and colleagues who found that aggregating environmental releases and activity levels across an entire sector without adjusting for inconsistent reporting has the potential to significantly underestimate U.S. average refinery releases.39 An alternate set of release factors for heat-related emissions and the FCC, toggleable by the user, were developed to reflect the most effective emissions controls in U.S. EPA’s AP-42 Emissions Database for these subprocesses (Table S25).40 Equation 1: Calculation of release factors. Adapted from Sengupta, Hawkins, and Smith.39 Release factor (F) of flow (i) from subprocess (s) is calculated from activity (A) and emissions (M) at refineries (n).

Table 3. Categorization of Subprocesses by Process Type for Release Factor Development process type heat-related process unit

support services miscellaneous managed wastes excluded

subprocess designation boilers, heaters, engines desalter, atmospheric distillation tower, fluid catalytic cracker, coker, residue hydrocracker, alkylation, asphalt unit, gas oil hydrocracker, sulfur plant, catalytic reformer, isomerization plant, hydrotreaters, steam-methane reformer flare, cooling tower, wastewater treatment, tank, terminal, fugitive emissions, blowdown systems aggregate facility releases to air not otherwise excluded releases from managed wastes chemical production, fluid coking, other subprocesses with inconsistent reporting

F′i , s =

(eq 1). This method ignores facilities that do not report a flow from that subprocess and thus represents the higher bound of the release factor. In the second approach (F″), total emissions of an elementary flow from a subprocess were divided by total activity (A′) of all facilities that report any flow from that subprocess in the relevant data set (eq 1). Under this method, nonreports of a flow are assumed to represent zero releases. In both cases, the release factors give greater weight to reported releases at higher-throughput facilities, which is appropriate given that they process a greater share of total crude. Releases of HAPs and toxins are more sensitive to the release factor calculation method as the reporting of these flows is less consistent across refineries. For these flows, the final release factor was calculated by weighting F′ and F″ according to the percent of total activity represented by the reporting facilities (eq 2). When reporting of CAPs and GHGs is consistent across a subprocess, F′ better represents the likely release rate and no weighted-average approach is needed. For release factors from heat-related combustion and the fluid catalytic cracker (FCC), the choice of release factor calculation affects total normalized impacts from that subprocess by ±2% and ±8%, respectively (Figures S13 and S16). Table 2 summarizes the source of release data and activity data for the development of each set of release factors. Additional details on the development of release factors for

Mi , s

F″i , s =

Ai , s

Mi , s A′s

Ai , s = Σnj aj , s for which mi , j , s exists A′s = Σnj aj , s for which mj , s exists n

mi , s =

∑ mi ,j ,s j

(1)

Equation 2: Calculation of weighted average release factor. Fi , s = Fi′, s

Ai , s As′

Ai , s zy ji zz + F″i , sjjj1 − z j As′ z{ k

(2)

The U.S. EIA Refinery Capacity data set reports refinery operating capacities and unit “charge capacities” for all U.S. refineries.27 We estimated utilization rates of operating capacity for each refinery, based on utilization rates of operable capacity reported by U.S. EIA for nine PADD regions,28 adjusting to account for idle capacity and refinery stream days by facility as reported in the Refinery Capacity data set27 and applying the adjusted utilization rate equally to the operating capacity of all refineries in the region (Table S9). Similarly, process unit activity levels were estimated from “downstream charge capacities” or “production capacities”, idle capacity and stream days by refinery, and sub-PADD D


Article


Figure 2. TRACI impacts of sector-wide refinery releases by category. “Total Normalized” reflects impacts normalized to 2008 U.S. impacts (Table S2).47 Impact categories are sorted in order of contribution to total normalized impacts. Heat-related and process unit releases are the primary drivers of normalized impacts from refineries. Sector wide totals do not include the “Excluded” subprocesses. Global warming potential (GWP), kg CO2 eq; photochemical ozone formation potential (POFP), kg O3 eq; acidification potential (AP), kg SO2 eq; particulate matter formation potential (PMFP), kg PM2.5 eq; eutrophication potential (EP), kg N eq; ozone depletion potential (ODP), kg CFC-11 eq.

Table 4. Comparison of Emissions Factors for Sources Found in PRELIM v1.2 PRELIM v1.2 combustion (g/MJ)a CO2 56.3 CH4 1.04 × 10−3 N2O 1.04 × 10−3 FCC catalyst coke combustion CO2 120.4 CH4 4.89 × 10−3 N2O 1.22 × 10−3 SMR (kg/m3) CO2 0.670

% change from v1.2

net change GWP

58.2 3.22 × 10−3 6.24 × 10−4 (kg/m3) 153.1 4.76 × 10−3 9.38 × 10−4

3.3% 210% −40%

3.3% 0.1% −0.2%

27% −2.7% −23%

27% 0% −0.1%

0.663

−1.1%

−1.1%

PRELIM v1.3

a

Figure 3. Contribution to global warming potential by source for selected crudes. Increase in GWP relative to PRELIM v1.2 is shown above each assay. U.S. Bakken (BK, hydroskimming), U.S. West Texas Sour (WTS, medium conversion), Saudi Arabian Light (SAL, medium conversion), and Canadian Cold Lake (CCL, deep conversion). These assays were some of the most common inputs to U.S. refineries in 201411 and reflect a range of refinery configurations, densities, and sulfur contents (Table S28).

Combustion emission factors in PRELIM v1.2 reflect natural gas, while PRELIM v1.3 represents the U.S. average mix of natural gas and refinery gas.

utilization rates (Table S10).31 The rate of idle capacity reported at a refinery is assumed to apply equally to all process units at that refinery as specific utilization rates of process units are not available. Consistent with U.S. EIA assumptions that utilization of major process units is similar to utilization of the crude distillation unit,41 the sub-PADD utilization rates were applied to all process units. Refineries were assigned a PRELIM configuration based on the presence of reported process units (Table S11) following the method described by Cooney and colleagues.11 U.S. EIA reports annual fuel consumption at refineries for each PADD (Table S13).29 Refinery gas and natural gas are used simultaneously and interchangeably in the same units within a refinery. As such, emissions reported from those units reflects the burning of both fuels. In this analysis, consumption of refinery gas and natural gas were combined to better reflect this fuel switching (Figure S11).31 Fuel consumption within each PADD was further allocated to individual

refineries proportional to utilized capacity. If a refinery reports any emissions from a given fuel type, in either the GHGRP or NEI, it was allocated a share of PADD-level fuel consumption proportional to its utilized operating capacity (Table S15). Validating Assignment of Releases to Subprocesses and Correcting for Outliers. Releases by process type were normalized to facility crude throughput to compare impacts across refineries (example for POFP shown in Figure 1). Differences across refineries may result from differences in fuel use, production efficiencies from scale, processing intensity, and emissions controls, among others. PRELIM is designed to represent typical refinery operations. Eighteen refineries, representing 0.7% of national operating capacity, were excluded from further analysis because insufficient data exist to characterize E


Article


Figure 4. 90% prediction intervals for TRACI impacts from refinery gas and natural gas combustion. Global warming potential (GWP), acidification potential (AP), particulate matter formation potential (PMFP), eutrophication potential (EP), ozone depletion potential (ODP), photochemical ozone formation potential (POFP). x-axis: billion MJ of natural gas and refinery gas consumption.

of that elementary flow across subprocesses (eq S1). This approach assumes that uncertainties of a given flow from multiple sources are independent. These prediction intervals are expressed in PRELIM as a percent of the expected release (e.g., Fi,s ± Pi,s%). Release factors less than zero are possible when prediction intervals exceed 100%. To ensure non-negative releases, aggregate uncertainties for each flow are approximated using a lognormal distribution (eq S2 and Figure S18). The parameters of the distribution are calculated such that the expected value of the log-normal distribution is set equal to the release factor, and the 95th percentile of the distribution is set to approximate the high-end range of the 90% prediction interval. Water Use. We estimated water withdrawals by process unit reflective of water lost to evaporation through the cooling tower, process water use, and process blowdowns resulting in wastewater generation (Figure S7 and Table S16). To estimate loss from the cooling tower, an evaporation factor of 1.68% was applied to estimates of cooling water use by

them (Table S18). Additionally, 8 instances of releases by specific refineries were excluded as outliers because their release rate was significantly higher than other refineries and responsible for a nontrivial contribution to total normalized impacts. The impact of these outliers is further discussed in Table S19. Uncertainty. We calculated the 90% prediction intervals for releases based on the regression of facility-level estimates for each elementary flow from each subprocess (Pi,s). The prediction intervals are evaluated as a function of the standard error of the prediction around the expected releases at the mean level of subprocess activity for U.S. refineries using eq 3. Equation 3: Calculation of prediction interval for subprocess release factors. Pi , s =

syx 1 + 1/n ·tcrit ŷ

(3)

For the releases by elementary flow for the unit process data set, we calculate the aggregate uncertainty as the square root of the sum of squares of the prediction intervals for instances F


Article


Figure 5. Total normalized impacts and global warming potential by refinery product for two assays in two refinery configurations. WTS, West Texas Sour; CCL, Canadian Cold Lake; Med, medium conversion coking refinery (3); Deep, deep conversion coking refinery (6); Jet, jet fuel; ULSD, ultralow sulfur diesel. Upstream and off-site managed waste impacts are included. Energy content allocation is used. Normalized impacts represent contributions by each impact category to total U.S. impacts per capita in 2008 (Table S2).47

process unit.42 Estimates of wastewater generation by process unit33 when applied to reported capacities align closely with the average refinery wastewater rate reported to the DMR of 0.8 L H2O/L crude (Figure S8). Water leaving the refinery as a constituent of products is assumed to be negligible.43 An evaporative cooling water system is assumed, although once-through and closed-loop systems also exist.43 The source for water withdrawal is assumed to reflect the average for U.S. refineries: 72% surface water, 10% groundwater, 18% municipal water.42 Total water withdrawal calculated in PRELIM is dependent on refinery configuration and crude assay but is within the range of estimates of freshwater withdrawals found in the literature, between 0.5−2.5 L H2O/L crude.44,45 Changes to the PRELIM Spreadsheets. The release factors described above were integrated into PRELIM v1.3 following the approach used for GHG releases in earlier versions of the model; release factors are applied to the activity level as calculated in PRELIM. Except for the FCC and steam methane reformer (SMR), releases from process units were not present in earlier versions of PRELM, as they were assumed not to significantly affect GWP. Similarly, support services such as flares, managed wastes, and other miscellaneous subprocesses

were not previously included. A total of 226 substances released across air, water, and soil are included in the updated PRELIM v1.3 model, accounting for 362 elementary flows and over 1800 subprocess−elementary flow combinations. As in earlier versions of the model, the expanded release inventories are calculated for each subprocess and allocated to refinery products according to user specifications. These relationships are summarized in an input output matrix added to PRELIM v1.3 to facilitate flow tracking and quality assurance. Releases from support services, miscellaneous sources, and managed wastes are allocated to refinery products based on the underlying crude flow or heat-demand by each product (Table S23). Full release inventories are provided per barrel of crude or as allocated to refinery products on a per MJ basis. The inventory is exportable to openLCA, an open-source LCA software,46 which enables rapid generation of refinery inventories as a component of larger LCA models or for use with alternate LCIA methods or characterization factors.

■

RESULTS AND DISCUSSION We were able to assign flows contributing to 98% of the total normalized impacts of the U.S. refinery sector. We assessed G


Article

Environmental Science & Technology impacts using TRACI v2.121 and normalized to 2008 U.S. national impacts (Table S2)47 to gauge the relative significance of the various flows (Table S20) and impact categories. Impact assessments for human health cancer, human health noncancer, and ecotoxicity are included within the PRELIM model but are not assessed in this study given the uncertainty in the underlying characterization factors from the USEtox model.48 We classified subprocesses as heat-related, process units, support services, miscellaneous, and managed wastes (Table 3) to understand their relative contributions to LCIA results. Releases from a small subset of subprocesses, such as production of products not tracked as outputs and fluid coking, were excluded because they are not included within PRELIM. These releases represented 1.9% of sector normalized impacts. Contributions by category are shown in Figure 2, and contributions from each subprocess are provided in the Supporting Information. Exploring TRACI impacts at refineries by process type reveals the range of expected impacts per unit of crude throughput and provides insight into differences in releases across refineries. In the case of POFP, higher release rates of nitrogen oxides and volatile organic compounds per unit of crude are associated with smaller refineries with less efficient combustion or fewer emissions control technologies (Figure S10). Differences across refineries in GWP per cubic meter of crude are driven more by the presence of additional process units than by efficiency of combustion. Refinery impacts per cubic meter of crude are more tightly bound for GWP, POFP, and AP than for PMFP and ODP (Figure S10). Validation of Release Factors. To gauge the representativeness of the new release data added to PRELIM v1.3, in Table 4 we compare releases of GHGs estimated using the new approach with the GHG estimates included in PRELIM v1.2. Prior work has validated the PRELIM GHG releases against other existing models.18 The carbon dioxide emissions factor we estimated for the combined combustion of natural gas and refinery is 3.3% greater than the carbon dioxide emission factor previously used in PRELIM v1.2 based on natural gas combustion. Our emission factor for methane is 210% greater than the previous PRELIM estimate, reflecting the lower efficiency of refinery gas combustion in our estimate. Our emission factor for nitrous oxide is 40% lower than the previous PRELIM estimate. However, the GWP impact of these changes to methane and nitrous oxide is minimal; on net, the updated emission factor for combustion is 3.2% greater than v1.2 for GWP. The differences between the prior PRELIM estimate and our updated estimates for releases from FCC catalyst coke combustion are reflective of the fact that our estimates are based on reported values for catalyst coke combustion while the previous values are based on coal combustion as a proxy. When emissions inventories are aggregated by assay at the refinery gate, the global warming potential per barrel of refined crude increases between 5.2% and 7.1% from version 1.2 (Figure 3). Most of this difference is explained by higher estimates of carbon dioxide release from FCC catalyst coke combustion and the inclusion of GHG releases associated with fugitive emissions, flaring, and other subprocess emissions not included in PRELIM v1.2. The release factors used in the PRELIM model represent the national average while acknowledging the variability in reported releases that likely reflect additional unaccounted for parameters at individual refineries. Prediction intervals are

used to better understand the variation in impacts from each subprocess. Impacts from heat-related combustion, which drive impacts for the most common crudes (Figure S19), are plotted in Figure 4 for the 128 refineries within the data sets. The prediction line reflects the forecasted impact as a function of fuel consumption. The width of the prediction interval around the data points indicates the level of confidence in the impacts per unit of fuel combusted. The tightest prediction interval is for GWP, while uncertainty is greatest for ODP. Model Results. Subprocess-based inventories developed for PRELIM v1.3 enable greater understanding of the drivers of impacts for refinery products across different assays. As a proof of concept, sample TRACI results are shown in Figure 5 for two assays (West Texas Sour [WTS] and Canadian Cold Lake [CCL]) processed in two different refinery configurations (medium and deep). The default refinery configuration for WTS is medium, while the default for CCL is deep. Both assay and configuration impact the contribution of refinery operations to normalized impacts. In particular, the additional refining steps used in a deep conversion refinery to convert liquid heavy ends to high-value products (i.e., gasoline, jet fuel, and ultralow sulfur diesel) results in higher impacts per MJ for these products under both assays. Additional comparisons of impacts between common assays are included in Figure S19. Model Sensitivity. A key feature of PRELIM is the ability to adjust refinery parameters and modeling choices. For example, PRELIM allows for the selection of allocation methods: mass, energy, hydrogen content, or market value basis. These methods alter the allocation of energy use and releases from subprocesses to intermediate flows. The sensitivity of refinery product impacts to the choice of allocation method differ across impact categories (Figure 6), depending

Figure 6. Sensitivity to impacts by product under alternative allocation. Graph shows impacts for each product by allocating subprocess demands to intermediate products using a hydrogen content basis relative to an energy content basis. Results represent Canadian Crude Lake assay at a deep conversion refinery (6).

on the driver of each impact. PMFP impacts for gasoline increase by 9% under a hydrogen allocation scenario relative to an energy allocation scenario for the assay represented in Figure 6. H


Article


Figure 7. Sensitivity to model parameters. Impacts per barrel of crude using Canadian Cold Lake assay at a deep conversion refinery (6).

Figure 8. TRACI impacts of gate-to-gate refinery emissions for gasoline, jet fuel (jet), and ultralow sulfur diesel (ULSD) refined from 2005 and 2014 U.S. petroleum mixes. Upstream releases are excluded (e.g., electricity and natural gas), while off-site managed waste releases are included. Hydrogen content allocation is used.

Other sources of model parametrization include options for refinery cogeneration and off-gas product production from gas treatment and desulfurization, which were existing features from PRELIM v1.2. In addition, new parametrization options include the ability to toggle on or off releases beyond the refinery gate, such as from upstream energy production or downstream waste management, and options for refinery emissions controls. The sensitivity to impacts per barrel of crude processed of these parameters are shown in Figure 7. Eliminating off-gas production of liquefied petroleum gases and petrochemical feedstocks increases net refinery fuel gas production, because less off-gas is refined into product. As a result, the refinery requires less external natural gas, which reduces impacts from upstream natural gas production. Excluding releases from off-site managed wastes has almost no effect on any of the impact categories included in this study, as these releases are most likely to impact toxicity-related impact categories. The net effects of the cogeneration feature within PRELIM, which replaces off-site electricity purchases

with on-site combustion, are mixed. Decreases in POFP, AP, GWP, and EP are offset by increase in PMFP and ODP. Any benefits of cogeneration are highly dependent on the default source of electricity selected by the user. 2005 to 2014 Impacts Comparison. The 2014 U.S. life cycle GHG petroleum baseline documents the changes in crude processed by refineries between 2005 and 2014 and the resulting changes in GWP.11 Figure 8 expands on that comparison by showing TRACI 2.1 impacts at the refinery for gasoline, jet fuel, and ultralow sulfur diesel (ULSD). Hydrogen allocation is used for this comparison, consistent with the petroleum baseline, as the hydrogen content of intermediate flows best reflects the processing intensity of each.11,15,49 As shown in the GHG petroleum baseline, GWP impacts increase for all fuels between 2005 and 2014, but the rest of the impacts show increases for gasoline and decreases for ULSD and jet fuel. Impact reductions at PADD 3 and PADD 5 drive the reductions for the U.S. average. Additional information and detailed results are included in the Table S29. I


Article


States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

In this study, an updated petroleum refinery model, PRELIM v1.3, is presented. The updated model provides an expanded, gate-to-gate inventory of environmental releases reflective of several user-defined parameters suitable for calculating a full suite of life cycle impact potentials using the TRACI or ReCiPe LCIA methods. Release factors for additional environmental flows at refinery subprocesses were formulated using a data mining approach on publicly reported refinery releases to air, water, and soil for 2014. Over 87% of normalized refinery impacts are assigned to specific refinery process units or heat production, with the remainder estimated at the facility level. The results of this effort provide significant opportunities to understand the life cycle impacts of transportation fuels and other refinery products beyond GHGs. This is especially important given the anticipated changes to the suite of crude assays entering U.S. refineries and helps identify potential environmental trade-offs. PRELIM v1.3 contributes to improving LCA analyses of petroleum products and will help inform policy makers of the environmental trade-offs associated with decisions affecting refinery operations.

■

■

(1) U.S. EPA. EPA Facility Level GHG Emissions Data. http:// ghgdata.epa.gov/ghgp/main.do (accessed Feb 13, 2018). (2) Young, B.; Krynock, M.; Carlson, D.; Hawkins, T. R.; Marriott, J.; Morelli, B.; Jamieson, M.; Cooney, G.; Skone, T. J. Comparative Environmental Life Cycle Assessment of Carbon Capture from Petroleum Refining, Ammonia Production, and Thermoelectric Power Generation. International Journal of Greenhouse Gas Control, submitted for publication, 2018. (3) Smargiassi, A.; Kosatsky, T.; Hicks, J.; Plante, C.; Armstrong, B.; Villeneuve, P. J.; Goudreau, S. Risk of Asthmatic Episodes in Children Exposed to Sulfur Dioxide Stack Emissions from a Refinery Point Source in Montreal, Canada. Environ. Health Perspect. 2009, 117 (4), 653−659. (4) McCoy, B. J.; Fischbeck, P. S.; Gerard, D. How Big Is Big? How Often Is Often? Characterizing Texas Petroleum Refining Upset Air Emissions. Atmos. Environ. 2010, 44 (34), 4230−4239. (5) Wilhelm, S. M. Estimate of Mercury Emissions to the Atmosphere from Petroleum. Environ. Sci. Technol. 2001, 35 (24), 4704−4710. (6) Diya’uddeen, B. H.; Daud, W. M. A. W.; Aziz, A. A. Treatment Technologies for Petroleum Refinery Effluents: A Review. Process Saf. Environ. Prot. 2011, 89 (2), 95−105. (7) Asghari, I.; Mousavi, S. M. Effects of Key Parameters in Recycling of Metals from Petroleum Refinery Waste Catalysts in Bioleaching Process. Rev. Environ. Sci. Bio/Technol. 2014, 13 (2), 139−161. (8) HSRC. Environmental Update #12: Environmental Impact of the Petroleum Industry. Hazardous Substance Research Centers and South & Southwest Outreach Program, June 2003. (9) U.S. EIA. Crude Oil Production; Energy Information Administration: Washington, DC, 2017. (10) Cooney, G.; Littlefield, J.; Marriott, J.; Skone, T. J. Evaluating the Climate Benefits of CO2-Enhanced Oil Recovery Using Life Cycle Analysis. Environ. Sci. Technol. 2015, 49 (12), 7491−7500. (11) Cooney, G.; Jamieson, M.; Marriott, J.; Bergerson, J.; Brandt, A.; Skone, T. J. Updating the US Life Cycle GHG Petroleum Baseline to 2014 with Projections to 2040 Using Open-Source Engineering-Based Models. Environ. Sci. Technol. 2017, 51 (2), 977− 987. (12) Barnes, B. US tight oil: Too light, too sweet. http://www. petroleum-economist.com/articles/markets/outlook/2017/us-tightoil-too-light-too-sweet (accessed Feb 16, 2018). (13) U.S. EIA. Annual Energy Outlook 2018: With Projections to 2050. U.S. Energy Information Administration, February 6, 2018. (14) Brandt, A. R. Variability and Uncertainty in Life Cycle Assessment Models for Greenhouse Gas Emissions from Canadian Oil Sands Production. Environ. Sci. Technol. 2012, 46 (2), 1253− 1261. (15) Wang, M.; Lee, H.; Molburg, J. Allocation of Energy Use in Petroleum Refineries to Petroleum Products. Int. J. Life Cycle Assess. 2004, 9 (1), 34−44. (16) Michalek, J. J.; Chester, M.; Jaramillo, P.; Samaras, C.; Shiau, C.-S. N.; Lave, L. B. Valuation of Plug-in Vehicle Life-Cycle Air Emissions and Oil Displacement Benefits. Proc. Natl. Acad. Sci. U. S. A. 2011, 108 (40), 16554−16558. (17) Elgowainy, A.; Han, J.; Cai, H.; Wang, M.; Forman, G. S.; Divita, V. B. Energy Efficiency and Greenhouse Gas Emission Intensity of Petroleum Products at U.S. Refineries. Environ. Sci. Technol. 2014, 48, 7612−7624. (18) Abella, J. P.; Bergerson, J. A. Model to Investigate Energy and Greenhouse Gas Emissions Implications of Refining Petroleum:

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.8b05572. Additional details on data, methods, and results (PDF)

■

REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. 626 Cochrans Mill Road, Pittsburgh, PA 15236-0940. ORCID

Ben Young: 0000-0001-6276-8670 Matthew Jamieson: 0000-0002-5853-9805 Gregory Cooney: 0000-0001-5853-5102 Kavan Motazedi: 0000-0002-1544-0485 Joule Bergerson: 0000-0002-4736-3509 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank John Guo, Ben Morelli, Sarah Cashman, Stacie Enoch, Eva Knoth, GreenDelta (Andreas Ciroth, Michael Srocka), Joe Marriott, David Morgan, and Timothy J. Skone for their support in model development, data processing, and manuscript review. This analysis was prepared by the office of Fossil Energy (FE) for the United States Department of Energy (DOE), National Energy Technology Laboratory (NETL). This study was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United J


Article

Environmental Science & Technology Impacts of Crude Quality and Refinery Configuration. Environ. Sci. Technol. 2012, 46 (24), 13037−13047. (19) Abella, J. P.; Motazedi, K.; Guo, J.; Cousart, K.; Bergerson, J. A. Petroleum Refinery Life Cycle Inventory Model (PRELIM). PRELIM v1.2 User Guide and Technical Documentation; University of Calgary, 2017. (20) Argonne National Laboratory. GREET, version 1_2016 Rev. 1; Argonne, IL, 2017. (21) U.S. EPA. Tool for Reduction and Assessment of Chemicals and Other Environmental Impacts, version 2.1; Environmental Protection Agency, 2014. (22) Huijbregts, M. A. J.; Steinmann, Z. J. N.; Elshout, P. M. F.; Stam, G.; Verones, F.; Vieira, M. D. M.; Hollander, A.; Zijp, M.; van Zelm, R. ReCiPe 2016: A Harmonized Life Cycle Impact Assessment Method at Midpoint and Endpoint Level Report I: Characterization. Int. J. Life Cycle Assess. 2017, 22 (2), 138−147. (23) U.S. EPA. Greenhouse Gas Reporting Program; Environmental Protection Agency, 2016. (24) U.S. EPA. 2014 National Emissions Inventory Data, version 2; Environmental Protection Agency, 2018. (25) U.S. EPA. Toxics Release Inventory Program; Environmental Protection Agency, 2016. (26) U.S. EPA. Discharge Monitoring Report Pollutant Loading Tool; Environmental Protection Agency, 2016. (27) U.S. EIA. Refinery Capacity Report; Energy Information Administration: Washington, DC, 2017. (28) U.S. EIA. Refinery Utilization and Capacity; Energy Information Administration: Washington, DC, 2017. (29) U.S. EIA. U.S. Fuel Consumed at Refineries; Energy Information Administration: Washington, DC, 2016. (30) Cashman, S. A.; Meyer, D. E.; Edelen, A. N.; Ingwersen, W. W.; Abraham, J. P.; Barrett, W. M.; Gonzalez, M. A.; Randall, P. M.; Ruiz-Mercado, G.; Smith, R. L. Mining Available Data from the United States Environmental Protection Agency to Support Rapid Life Cycle Inventory Modeling of Chemical Manufacturing. Environ. Sci. Technol. 2016, 50 (17), 9013−9025. (31) Young, B.; Morelli, B.; Hawkins, T. R. Creation of Unit Process Data for Life Cycle Assessment of Steam Methane Reforming and Petroleum Refining; Argonne National Laboratory, 2017. (32) Meyer, D. E.; Abraham, J. P. Source Reporting Requirements for Life Cycle Inventory Modeling Using EPA Data Sources: Examples for the Rapid Chemical Life Cycle Inventory Modeling Method, draft report; U.S. Environmental Protection Agency: Cincinnati, OH, 2017. (33) U.S. EPA. Technical Support Document for the 2004 Effluent Guidelines Program Plan; EPA-821-R-04-014; U.S. Environmental Protection Agency: Washington, DC, 2004. (34) Fu, F.; Wang, Q. Removal of Heavy Metal Ions from Wastewaters: A Review. J. Environ. Manage. 2011, 92 (3), 407−418. (35) U.S. EPA. Factors to Consider When Using Toxics Release Inventory Data; Environmental Protection Agency, 2015. (36) Marafi, M.; Stanislaus, A. Spent Catalyst Waste Management: A Review: Part IDevelopments in Hydroprocessing Catalyst Waste Reduction and Use. Resour. Conserv. Recycl. 2008, 52 (6), 859−873. (37) Akcil, A.; Vegliò, F.; Ferella, F.; Okudan, M. D.; Tuncuk, A. A Review of Metal Recovery from Spent Petroleum Catalysts and Ash. Waste Manage. 2015, 45, 420−433. (38) Dempsey, C. R.; Oppelt, E. T. Incineration of Hazardous Waste: A Critical Review Update. Air Waste 1993, 43 (1), 25−73. (39) Sengupta, D.; Hawkins, T. R.; Smith, R. L. Using National Inventories for Estimating Environmental Impacts of Products from Industrial Sectors: A Case Study of Ethanol and Gasoline. Int. J. Life Cycle Assess. 2015, 20 (5), 597−607. (40) U.S. EPA. WebFIRE; Environmental Protection Agency, 2017. (41) U.S. EIA. Refinery Outages: First-Half 2016; Energy Information Administration: Washington, DC, 2016. (42) Jacobs Consultancy. Water Consumption in US Petroleum Refineries; 2016. (43) IPIECA. Petroleum Refining Water/Wastewater Use and Management; London, United Kingdom, 2010.

(44) Wu, M.; Mintz, M.; Wang, M.; Arora, S. Water Consumption in the Production of Ethanol and Petroleum Gasoline. Environ. Manage. 2009, 44 (5), 981−997. (45) Laurenzi, I. J.; Bergerson, J. A.; Motazedi, K. Life Cycle Greenhouse Gas Emissions and Freshwater Consumption Associated with Bakken Tight Oil. Proc. Natl. Acad. Sci. U. S. A. 2016, 113 (48), E7672−E7680. (46) Ciroth, A. ICT for Environment in Life Cycle Applications OpenLCAA New Open Source Software for Life Cycle Assessment. Int. J. Life Cycle Assess. 2007, 12 (4), 209. (47) Ryberg, M.; Vieira, M. D.; Zgola, M.; Bare, J.; Rosenbaum, R. K. Updated US and Canadian Normalization Factors for TRACI 2.1. Clean Technol. Environ. Policy 2014, 16 (2), 329−339. (48) Bare, J. TRACI 2.0: The Tool for the Reduction and Assessment of Chemical and Other Environmental Impacts 2.0. Clean Technol. Environ. Policy 2011, 13 (5), 687−696. (49) Bredeson, L.; Quiceno-Gonzalez, R.; Riera-Palou, X.; Harrison, A. Factors Driving Refinery CO2 Intensity, with Allocation into Products. Int. J. Life Cycle Assess. 2010, 15 (8), 817−826.

K


Expansion of the Petroleum Refinery Life Cycle Inventory Model to

Recommend Documents