This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.
Article pubs.acs.org/est
Updating the U.S. Life Cycle GHG Petroleum Baseline to 2014 with Projections to 2040 Using Open-Source Engineering-Based Models Gregory Cooney,† Matthew Jamieson,† Joe Marriott,† Joule Bergerson,‡ Adam Brandt,§ and Timothy J. Skone*,† †
National Energy Technology Laboratory, 626 Cochrans Mill Road, P.O. Box 10940, Pittsburgh, Pennsylvania 15236, United States University of Calgary EEEL Building University of Calgary, 2500 University Drive NW, Calgary, Alberta Canada T2N 1N4 § Stanford University, 066 Green Earth Sciences Building, 367 Panama St., Stanford, California 94305, United States ‡
S Supporting Information *
ABSTRACT: The National Energy Technology Laboratory produced a well-to-wheels (WTW) life cycle greenhouse gas analysis of petroleum-based fuels consumed in the U.S. in 2005, known as the NETL 2005 Petroleum Baseline. This study uses a set of engineering-based, open-source models combined with publicly available data to calculate baseline results for 2014. An increase between the 2005 baseline and the 2014 results presented here (e.g., 92.4 vs 96.2 g CO2e/MJ gasoline, + 4.1%) are due to changes both in modeling platform and in the U.S. petroleum sector. An updated result for 2005 was calculated to minimize the effect of the change in modeling platform, and emissions for gasoline in 2014 were about 2% lower than in 2005 (98.1 vs 96.2 g CO2e/MJ gasoline). The same methods were utilized to forecast emissions from fuels out to 2040, indicating maximum changes from the 2014 gasoline result between +2.1% and −1.4%. The changing baseline values lead to potential compliance challenges with frameworks such as the Energy Independence and Security Act (EISA) Section 526, which states that Federal agencies should not purchase alternative fuels unless their life cycle GHG emissions are less than those of conventionally produced, petroleumderived fuels.
■
INTRODUCTION The National Energy Technology Laboratory (NETL) produced a well-to-wheels life cycle greenhouse gas (GHG) analysis of petroleum-based transportation fuels consumed in the United States in 2005, known colloquially as the NETL 2005 Petroleum Baseline.1 The boundaries of that study included production of crude oil through final combustion of petroleum fuel products. In 2007, the Energy Independence and Security Act (EISA) required the EPA via the Renewable Fuel Standard (RFS2) to establish a life cycle baseline to measure renewable fuels against. EPA selected the NETL Petroleum Baseline analysis as the basis for life cycle GHG emissions of conventionally produced petroleum-based transportation fuels.2,3 EISA Section 526 states that federal agencies should not purchase alternative fuels unless the life cycle GHG emissions are less than those of conventionally produced, petroleum-derived fuels. Aside from federal procurement, the NETL Petroleum Baseline has also been cited extensively in the literature as the reference source for GHG emissions for the transportation fuel supply chain. Crude production has changed substantially since 2005 baseline. There has been a shift in the sources of crude oils processed and consumed in the U.S., due primarily to the tight © XXXX American Chemical Society
oil boom. Domestic crude oil production has increased 68% compared to 2005 (5.2 million bbls/day to 8.7 million bbls/day in 2014).4 The largest increases in production have come in the states of North Dakota (Bakken Shale) and Texas (Eagle Ford Shale) at 1 and 2 million bbls/day, respectively. In 2014, for the first time in several decades, domestically produced crude overtook imports as the majority share of oil processed by U.S. refineries. Even though total imports have fallen from 10 to 7.3 million bbls/day from 2005 to 2014, imports from Canada have increased by over 50%.5 More than half of the growth in imports from Canada has come in the form of diluted bitumen and synthetic crude oil extracted and upgraded from the Canadian oil sands.6 There have also been major changes to the refining sector; one was the shift from low-sulfur (500 ppm of S) to ultra-lowsulfur diesel (15 ppm of S) in 2006.7 This change resulted in increased hydrogen requirements for hydrotreating processes. The growth in the domestic production of light and sweet tight Received: June 6, 2016 Revised: October 19, 2016 Accepted: November 3, 2016
A
DOI: 10.1021/acs.est.6b02819 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
CO2e/MJ delivered) and the U.S. average grid mix for electricity (609 kg CO2e/MWh delivered).25,26 They were also updated to utilize the latest IPCC factors for Global Warming Potential based on the Fifth Assessment Report (AR5).27 This analysis utilizes the latest publicly available versions of OPGEE (v1.1 Draft E) and PRELIM (v1.0). Production GHG emissions for Canadian oil sands are modeled using the GreenHouse gas emissions of current Oil Sands Technologies (GHOST) model. GHOST quantifies the GHG emissions associated with the production of bitumen via surface mining and in situ methods as well as the upgrading of the produced bitumen to synthetic crude oil.28,29 The emissions associated with the production of crude by CO2 Enhanced Oil Recovery (EOR) with CO2 sourced from natural domes were adapted from modeling work produced by NETL.30,31 Data Requirements and Availability. The complexity of these tools results in considerable data requirements to fully characterize the modeled systems. OPGEE requires the population of nearly 60 parameters to fully describe the production, surface processing, and transport emissions for a particular crude oil. OPGEE includes a set of default values so that results can be computed even if a user cannot fully populate the inputs. Previous studies have indicated that there is a subset of key parameters (recovery method, API gravity, field depth, gas-oil ratio, water−oil ratio, steam−oil ratio, and flaring rate) that when fully populated can reliably describe a particular conventional crude oil.15,16,20,22,23,32 However, data for even just these key parameters are not readily available, especially for crude oils produced internationally. PRELIM utilizes five parameters to characterize the quality of the whole crude: crude distillation curve, sulfur content, API gravity, carbon residue content, and hydrogen content. The five quality parameters are specified for each of nine crude fractions in addition to the whole crude, resulting in a total of 62 required parameters for each crude. This data is generally included in a crude oil assay. While a limited subset of crude assays are available on oil producer’s Web sites, other assays are considered proprietary. The public version of PRELIM contains 65 unique crude oil assays. As part of this effort, an additional 17 assays were added to more fully encompass the source countries for U.S. imports (SI Table SI-4). PRELIM also requires inputs related to the operations of the refinery (e.g., operating temperatures, pressures, hydrogen consumption) and is preloaded with default values. Crude Oil Sources and Quality. Proprietary data restrictions limit the ability to fully characterize all crude oils produced and consumed in the U.S., so this analysis instead models a smaller basket of crude oils to represent the consumption mix (SI Table SI-4). Brandt et al. determined that the uncertainty in a sample of crudes is generally much less than the uncertainty in a particular crude, for any level of data gathering effort due to compensating errors.33 This implies that despite some data gaps, the inherent uncertainty is deemed reasonable. To effectively utilize the models, it is necessary to understand the source, quantity, and quality of crude oil imported to and produced within the U.S. The primary source for this information throughout the entirety of this study is data collected by the U.S. Energy Information Administration (EIA). Imports. The EIA tracks the imports of all petroleum products that enter the U.S. via Form EIA-814 − CompanyLevel Imports.5 For each import, the database includes the source country, the import PADD, processing PADD, processing company, import quantity, and crude quality
oil has changed the overall quality of crude consumed in the U.S. This has presented a challenge to refineries which have largely adapted to process heavier, sourer crude oils.8−10 It has also had the effect of reducing imports of light/sweet crude oils from countries like Nigeria. Until recently, exports of crude oil from the U.S. had been prohibited for the last four decades.11,12 As a result, the API gravity (an inverse measure of density) of crude processed in the U.S. has increased by 4% since 2005 from 30.2 to 31.4, while the average sulfur content remained relatively flat (1.42 wt % S in 2005 and 1.45 wt % S in 2014).13 As part of the Consolidated Appropriations Act signed on December 18, 2015, the export ban was lifted and crude oil was reclassified such that no license is required for exporting.14 Life cycle GHG emissions associated with the production, refining, and combustion of petroleum-based fuels have been studied extensively in the literature (see the Supporting Information (SI) for more information).1,15−20 Some of the existing studies have relied on proprietary data and models to generate results. In others, the scopes of the analyses were limited to specific crude oil comparisons (e.g., oil sands vs conventional), regional evaluations (e.g., State of California), or non-U.S. specific pathways (e.g., European Union). This study calculates results that are representative of petroleum fuels produced and consumed in the U.S. in 2014 with the goal of establishing a 2014 baseline, and quantifying in GHG terms, the changes described above. In addition to the national level, this analysis examines differences in the well-to-wheel emissions at the Petroleum Administration for Defense District (PADD) level (see SI Figure SI-4). The results from this analysis are compared to the NETL 2005 Petroleum Baseline.1 A complementary objective is developing a framework that utilizes an engineering-based approach to calculating the life cycle GHG emissions with open-source and fully transparent models and data where available. This framework makes it possible to analyze potential policies in future studies for reducing carbon emissions from the full system or particular portions, and to see the effect of changes within the life cycle. This framework also enables the forecasting of future changes to the environmental profile of fuels produced in the U.S. Analyzing long-term projects (e.g., investments for compliance with EISA Sec. 526) against a forecasted baseline when a technology or policy change is implemented can improve the understanding of potential benefits over a single historic baseline value.
■
MATERIALS AND METHODS Modeling Framework. This analysis relied upon two open-source tools to model the energy and GHG emissions from the production and refining of crude into finished products. The Oil Production Greenhouse gas Emissions Estimator (OPGEE) is an engineering-based model that estimates GHG emissions from the production, processing, and transport of crude oil.21−23 OGPEE can model primary through tertiary production of crude oil and can be tuned to represent variations in production practices and geology. The Petroleum Refinery Life Cycle Inventory Model (PRELIM) is a detailed mass and energy based representation of the refining process that allows for the estimation of GHG emissions.24 PRELIM provides the flexibility to model up to ten different refinery configurations moving from simple (hydroskimming) to complex (deep conversion). OPGEE and PRELIM were modified to utilize NETL life cycle GHG values for the modeling of upstream utility inputs of natural gas (14.8 g B
DOI: 10.1021/acs.est.6b02819 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
A blending optimization procedure was created with the objective of minimizing the differences between the EIAreported and modeled crude quality data based on the available assays for each country. The absolute values of the percent deviations from the targeted API and sulfur were calculated and summed and served as the objective in the blending optimization process. In a 2014 study, Elgowainy et al. constructed a linear regression model to predict the overall efficiency of a refinery based on the API, sulfur, percentage of heavy products yield, and refinery complexity index.35 The refinery efficiency was defined as the ratio of the total energy associated with the refinery products and the sum of all energy inputs to the refinery including feedstock. That expression was utilized to determine the sensitivity to overall refinery efficiency due to changes in API and sulfur. Overall refinery efficiency is more sensitive to API than sulfur by a factor of 3.29. Assuming that refinery efficiency is a reasonable proxy for the overall GHG intensity of the refinery operations, this factor was applied to the calculated API in the blending optimization scheme. Utilizing this method, the mix of crudes from a single country consumed within a given PADD were optimized to minimize the total deviation. This procedure was repeated for each PADD and each of the countries with multiple assays. The same procedure was repeated for the entire U.S. Since this analysis relies on a smaller basket of crude oils to represent the consumption total U.S. consumption mix, in some cases, the volume demanded from a particular field is larger than the total production from that field. That limitation is a function of the data availability for OPGEE and PRELIM and the assumption is that the crude mix from that country can be effectively modeled as looking like a particular mix of a small subset of fields. The optimization procedure is described by the following equations:
(API/sulfur). This is the only data set available from EIA that includes crude quality data by country. The primary countries of interest considered in this study accounted for over 95% of all of the crude oil imported by the U.S. in 2014: Canada (39.3%), Saudi Arabia (15.8%), Mexico (10.6%), Venezuela (10.0%), Iraq (5.0%), Kuwait (4.2%), Colombia (4.0%), Ecuador (2.9%), Brazil (2.0%), Angola (1.9%), Nigeria (0.8%), and Russia (0.2%) (SI Table SI-2). In addition to the contributions to the U.S. mix, these countries were selected because data exists to effectively model them in OPGEE and PRELIM. None of the countries excluded from this analysis contribute more than 1.0% to the U.S. import mix. For each country, the 2014 volume-weighted average API and sulfur were calculated for each PADD (see SI Table SI-4, Figure SI-4). Domestic Crude Production. EIA provides information on the annual domestic production of crude oil by state and PADD; however, it does not track any quality data.4 EIA also publishes data on movements of crude between PADDs.34 It was inferred that these crude movements are of domestic, not foreign crude. The company level imports data set indicates that all imports are consumed in the PADD to which they are imported. The difference between the total crude movements from a PADD and the amount of crude produced in that PADD is assumed to be processed in that PADD. The purposes of evaluating this data is to generate a mix of domestic crudes that are consumed in a given PADD, by source PADD (see SI Tables SI-5 and SI-6). Crude Assay Blending. As shown in SI Table SI-4, some countries export products of varying quality and have several available assays (e.g., Canada at 21 and Nigeria at 8), while others have a limited set (e.g., Colombia, Iraq, Ecuador, Russia all at 1 each). Therefore, it was necessary to determine the appropriate blend of assays for the following countries: Angola, Brazil, Canada, Kuwait, Mexico, Nigeria, Saudi Arabia, and Venezuela. The purpose of the blending process was to match the API and sulfur reported for the imports from each of the target countries to a mixture of the available assays in PRELIM. For a given country, the “target” API (calculated from the specific gravity) and sulfur were calculated from the EIA Company Level Imports Data. calculatedAPI =
⎤ ⎡ calculatedAPI calculated sulfur min⎢ − 1 × 3.29 + −1 ⎥ ⎥⎦ ⎢⎣ targetAPI target sulfur
141.5 − 131.5 (SpGrCrude1 × vol fraction Crudei + ... + SpGrCruden × vol fraction Cruden)
percentage of each crude type consumed in a given PADD was added as an additional objective function in the optimization model. By utilizing the in-country assay blends, an import blend was created for each PADD and the corresponding quality (API/ sulfur) was determined. It was possible to calculate the API and sulfur of domestically produced crude by PADD since the quality/quantity of imported crude was known, along with the quality/quantity of total crude processed in the U.S., which is reported by EIA. 4,5,13 PRELIM contains crude quality information on 12 different U.S. crudes, originating from PADDs 2, 3, and 5. Crude blends were created for PADDs 1 and 4 based on averages of other assays to match API and sulfur. The U.S. blend for each PADD had a target API and sulfur. There were also targets for the percentage of crude for a selected PADD that was sourced from each of the other PADDs.
calculated sulfur = sulfurCrude1 × Vol fraction Crudei + ... + sulfurCruden × vol fraction Cruden
vol_fraction Crudei ≤ 1 n
∑ vol_fractionCrudei = 1 i=1
The procedure for blending the Canadian crudes was modified to account for additional available information regarding the type of crude available from the Canadian National Energy Board (NEB). Each of the available Canadian crudes is designated as conventional (light [>30 API]/ med[25−30 API]/heavy[
