Expected final online publication date for the Annual Review of Chemical and Biomolecular Engineering , Volume 15 is June 2024. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.
]]>The chemical industry plays a vital role in modern society. In 2017, the industry contributed $5.7 trillion (around 7.1%) to global gross domestic product and provided 120 million jobs (1). The chemical industry is one of the largest energy consumers and greenhouse gas (GHG) emission sources. In 2020, global chemical production consumed 9% of natural gas and 13% of oil and emitted 13% of industrial direct CO 2 emissions (2)(3)(4). Decarbonization of the chemical industry is needed urgently, especially given the critical role it may play in supporting the net-zero targets of other sectors, e.g., supplying low-carbon fuels for energy and transportation.
Several global studies have explored decarbonization pathways for the chemical industry (2,(5)(6)(7)(8)(9). Other studies have reviewed decarbonization technologies for chemical production (10)(11)(12)(13). Previous literature has provided quantitative assessment and data using various modeling approaches, such as process analysis, material flow analysis (MFA), life cycle assessment (LCA), techno-economic analysis (TEA), and machine learning (ML). Those methods have been reviewed individually for specific pathways, e.g., process analysis for plastic recycling (14), or for chemicals in general (15,16). As a supplement and update to previous reviews, this article analyzes how different analytical methods have been advanced and integrated for assessing decarbonization technologies and strategies in the chemical industry, as well as how chemical engineers can leverage industrial ecology and economic tools to better understand and quantify systems-level effects of technology deployment and industrial decarbonization. This article focuses on the models that not only assess the decarbonization potential of various pathways but also evaluate the technical, environmental, and economic implications of implementing decarbonization technologies and strategies in chemical production.
There are many decarbonization pathways for the chemicals industry. This review focuses on promising pathways identified by previous studies (2,13), including carbon capture and storage (CCS) and carbon capture and utilization (CCU), alternative feedstocks, and electrification and other process improvements. This study also includes demand-side strategies, such as circular economy practices that enhance resource utilization efficiency and reduce chemical demand. The environmental, economic, and decarbonization performances of technologies often differ by scale. The multi-scale effects must be assessed by different analytical methods (see Figure 1). Therefore, this review first introduces analytical methods (Section 2), followed by their applications for micro-/mesoscale (Section 3) and macroscale (Section 4) assessments of decarbonization technologies in the chemical industry.
This section briefly introduces five methods that previous studies have applied for evaluating decarbonization technologies and strategies in the chemical industry. MFA and LCA are industrial ecology methods that have been applied to broad industrial activities and products. The recent MFA and LCA applications for chemicals have leveraged chemical process analysis and TEA to enable the assessment and optimization of emerging or traditional technologies considering technical, environmental, and economic performance. Among the five methods reviewed here, ML is the only data-driven method and has been applied to chemical processes when knowledge-based information is lacking. This section briefly introduces the general concepts and examples of these methods, and their applications to individual decarbonization pathways in the chemical industry are reviewed in Section 3.
Yao et al. Micro-, meso-, and macro-level modeling techniques for decarbonization technologies and strategies in the chemical industry (17)(18)(19)(20)(21).
MFA is a key methodology in the study of the physical functioning of society. The goal of an MFA is to quantify the input and output flows of a specific material as it proceeds through its life cycle from initial mineral deposit acquisition to processing, incorporation into intermediate products (such as resins), manufacture of final goods (such as piping), incorporation into use, and (eventually) recovery for reuse or (in many cases) discard.
Figure 2 illustrates a typical (although complex) MFA: plastics in the United States in 2015 (22). Seven specific plastics and one plastics group are followed quantitatively step by step from
LCA is a standardized method to evaluate environmental impacts across the life cycle of a product or service, including raw material acquisition, production, use phase, and end of life (cradle to grave) (25). LCA includes four phases: goal and scope definition, life cycle inventory (LCI) analysis, life cycle impact assessment, and interpretation. During the goal and scope definition phase, the purpose of an LCA (e.g., intended use and audience of the study) and scope (e.g., functional unit, system boundary, environmental impact categories) must be defined and determined clearly. During LCI analysis, relevant inputs and outputs of processes (e.g., mass and energy balances) within the system boundary are collected. The compiled LCI results are converted to specific environmental impacts in the life cycle impact assessment phase. Finally, the results are interpreted based on the goal and scope of the study, and necessary sensitivity checks and quality evaluations are conducted before conclusions and recommendations are made. Kleinekorte et al. (16) and Bakshi (15) have reviewed LCA applications for the sustainable design of chemical products, processes, and supply chains.
Carbon accounting and carbon footprint analysis are based on LCA methods but focus only on GHG emissions and sinks. Different GHGs, such as CO 2 , CH 4 , N 2 O, and fluorinated gases, often are aggregated based on their global warming effects across a time horizon relative to CO 2 and reported in the unit of CO 2 equivalent (26). The Intergovernmental Panel on Climate Change (26) publishes characterization factors of non-CO 2 GHGs for different time horizons (e.g., 20, 100, and 500 years). Carbon accounting can be applied at product, process, corporation, landscape, national, and global levels and often classifies GHG emissions into three scopes. Scope 1 includes direct, on-site GHG emissions; Scope 2 includes GHG emissions associated with purchased electricity; and Scope 3 GHG emissions are related to upstream (e.g., material acquisition) and downstream (e.g., end-of-life) operations (27).
Process analysis in the chemical industry analyzes the process performance of chemical production, e.g., investigating material and energy balances, process variables, and properties of process materials (28). Because many decarbonization technologies have low technology readiness levels (TRLs), industrial-scale process information (e.g., mass and energy balances, environmental emissions) needed by LCA and TEA is often lacking. Process analysis provides engineering-rigorous assessments of potential large-scale production based on lab-verified data and/or process simulations, offering a robust means to address the data challenges of TEA and LCA. For low-TRL systems with alternative reaction and process pathways, superstructure-based process design, synthesis, and optimization methods have been developed in process systems engineering, which generates process flowsheets with conditional flow paths and computes an optimal structure for one or multiple objectives, e.g., to maximize or minimize specific performance indicators (29). As Figure 1 shows, process analysis can scale up conceptual/lab-scale exploration to a computational level relevant to real-world chemical production and supply chains.
TEA evaluates the economic performance of a product, process, or service based on its technical performance (21). TEA has been used widely to assess the economic viability of emerging technologies and products, commonly at the process or plant level. A TEA framework similar to the four phases of LCA has been proposed (21). In the phase of goal and scope definition, the overall goal is defined along with the determined scope of the TEA. In the inventory phase, the technical data (e.g., mass and energy balances, process variables or parameters) and economic data (e.g., equipment cost, labor cost) are collected and analyzed. In the calculation phase, technical and economic performance results are generated based on selected indicators (e.g., minimum selling price, net present value, energy efficiency). And in the interpretation phase, the technical and economic performance results are interpreted based on the goal and scope of the TEA study (21).
ML is defined as a computer program that "is said to learn from experience (E) with respect to some class of tasks (T ) and performance measure (P), if its performance at tasks in T, as measured by P, improves with experience E" (30, p. 2). Common ML techniques include artificial neural networks, support vector machines, and random forest (31). In a typical ML workflow, a model is trained and validated using data collected from experiments or literature and then used for prediction or optimization. ML has been used to develop surrogate models for chemical processes or production systems (e.g., based on process simulation data or operating data), optimization, process control, and fault diagnosis (31,32). Several studies have reviewed ML applications in the chemical industry (31,32). Therefore, this review includes mainly ML applications that enhance MFA, LCA, TEA, and process analysis of decarbonization technologies in the chemical industry.
This section discusses micro-and mesoscale modeling for three decarbonization pathways in the chemicals industry: carbon capture, utilization, and storage (CCUS); alternative feedstock; and electrification and other process improvements. Microscale modeling simulates unit operations or a chemical plant equipped with one or multiple decarbonization technologies. Mesoscale modeling includes interconnected manufacturing activities or groups, such as the supply chain or life cycle of a chemical, or chemical plant networks.
CCUS refers to a suite of technologies involving the capture of CO 2 from diverse sources (carbon capture, CC), storing CO 2 in underground geologic formations (carbon storage), or the chemical transformation of CO 2 into a product (carbon utilization) (33). CO 2 can be sourced from industrial facilities or directly from the air. Different CC strategies exist, including postcombustion, precombustion, and oxy-fuel combustion. Common techniques to capture/separate CO 2 include chemical/physical absorption and adsorption, and ionic liquids (ILs) and hybrid systems are emerging (34). CO 2 can be transported by pipeline, ship, rail, or truck. Deep geological formations for storing CO 2 include depleted oil and gas reservoirs and saline reservoirs. CCS provides long-term CO 2 storage, and the best practices for monitoring and verification to ensure the storage permanence have been discussed widely (2,33). For CCU systems, the fate of CO 2 and climate change mitigation benefits highly depends on CO 2 utilization options (35). For instance, CCU for long-lived products (e.g., construction materials) can provide relatively longterm carbon storage and achieve net carbon negative if CO 2 is from biogenic sources or direct air/water capture; on the contrary, carbon in CCU for short-lived products (e.g., fuels and chemicals) will be released to the atmosphere in a short time. However, CCU provides opportunities to replace fossil carbon with existing or new products and chemicals, enabling a circular carbon economy that brings additional social, economic, and environmental benefits that CCS alone cannot achieve. Chemical plants for carbon-intensive commodity chemicals (e.g., olefins, methanol, aromatics, ammonia and derivatives) are often considered for CCS implementation, given their GHG-intensive production processes (these chemicals account for >70% of global GHG emissions from the chemical industry) (2). The carbon sources of CCU can be from the chemical industry or other sectors (e.g., the power sector). The chemical products from CCU have large variations, but most of them are still in fundamental research. Commercial CCU products are very limited; a few examples include methanol, dimethyl ether, carbon monoxide, and some polymers (33). Both CCS and CCU will be needed to decarbonize the chemical industry and support a net-zero society.
In studying any material flow system, the first step is to identify and quantify mass flows, and thus reveal the most important components. Figure 3 provides such an analysis for the entire global chemical and petrochemical sector in 2013. The Sankey diagram immediately shows the relative flows of the feedstocks, the upstream and downstream chemicals, and the principal chemical products. Given these perspectives, it becomes straightforward to develop approaches to decarbonize the chemical industry while considering control aspects such as CCUS (2,7,8,36).
Decarbonizing the chemical/petrochemical sector will require that the system not only captures carbon but also develops feasible and affordable approaches to reuse it (7,37,38). Doing so at the required scale is an enormous challenge in technology development, industrial participation, and global collaboration. Previous studies have used MFA to track carbon flows within one chemical plant or across chemical networks, enabling the identification of main GHG emission sources, and pairing carbon sources and sinks for effective decarbonization. For example, Yao et al. (39) analyzed GHG flows in a representative petroleum refinery to estimate the cost and mitigation potential of different CC implementation strategies, and Kaiser & Bringezu (40) modeled carbon flows for material use in Germany to evaluate the potential of CO 2 utilization to close the carbon cycle of the chemical industry.
An important but often overlooked aspect of decarbonization of the economy is that such a transition must address at least two potentially confounding issues: (a) carbon in products such as cosmetics and coatings that often dissipate or disintegrate during use ( 41) and (b) commercial goods and packaging that are often lost during or after use (41), thus making a fully circular chemical industry very difficult. Efforts to surmount these challenges will thus be needed (22,24).
Process simulation and other chemical engineering computer-aided tools have been explored in microscale process analysis, e.g., for pilot plant or industrial-scale operations (42). Roh et al. (42) presented a systematic procedure for using these tools, including screening CCU routes using stoichiometry-based methods, calculating mass and energy balances using flowsheet-based process simulators (e.g., Aspen Plus, HYSIS, CHEMCAD), and leveraging primary data collected from lab-scale experiments. Their study highlighted that the accuracy of these models for low-TRL CCUS depends strongly on process design and performance that often rely on theoretical evidence or expert judgment.
All values in million tonnes per year (Mt yr Many process systems engineering studies have developed superstructure-based models for exploring alternative designs and addressing uncertainties for CCUS system synthesis and optimization (29). One example is ArKaTAC 3 , a tool to explore combinations of carbon sources, CC process options, and final products and optimize the superstructure for maximum profits and net CO 2 reduction using mathematical programming (43). Other studies have linked microlevel models of CCUS systems with carbon sources (e.g., fossil power plants) or with meso-level models for supply chain management across CC, transport, and storage/utilization at various temporal and spatial scales (44). These linked models are helpful in assessing and optimizing large-scale CCUS deployment and supporting policy and industrial decision making. In contrast, other studies have linked process-level models of CCUS systems with smaller-scale (e.g., molecule-level) models for screening and optimizing CC materials (e.g., absorbents and adsorbents), which are helpful for research and development (44). Multi-scale models are computationally challenging, given the high computational costs and extensive knowledge required to develop and use process models.
ML has been deployed to address computational and knowledge challenges. Microlevel ML applications allow for a sequential, bottom-up approach to screen many materials and technologies before conducting a process analysis (44). For instance, Wang et al. (45) screened different ILs based on ML-predicted viscosity and melting point and developed process simulations only for the best ILs selected by ML (45). Similar applications for different CC systems have been reviewed (46). Furthermore, ML supports interlinked micro-and meso-models of CCUS networks by providing computationally cheaper surrogate alternatives that address the challenges in solving nonlinear algebraic and partial differential equations for optimizing dynamic, cyclic CC processes (e.g., pressure swing adsorption) (44) or constructing input-output relationships when data and knowledge gaps exist (47). ML training data can be measured data (48) or come from process simulations (e.g., Aspen Plus or gPROMS) (47). Most ML applications in process analysis are limited to the individual life cycle stage of CCUS (e.g., CC or CO 2 utilization). More universal workflows for ML applications across the CCUS life cycle and development of hybrid knowledge-based/data-driven models are needed to facilitate a holistic assessment of CCUS.
Techno-economic analysis. CO 2 removal and avoidance costs are two common performance indicators that were widely assessed in previous microscale TEA models for CCS systems. CO 2 removal cost is calculated as the increased production cost divided by removed CO 2 (49); CO 2 avoidance cost is the increased production cost divided by reduced CO 2 emissions compared to chemical production without CCS (49,50). Because CCU systems generate marketable products, previous CCU TEAs have focused on evaluating product-oriented and profit-relevant indicators, such as internal rate of return, net present value, capital investment, discounted payback period, and minimum selling price of CO 2 -drived chemicals (51). Besides microlevel analysis, TEA models have been coupled with optimization models to evaluate the mesoscale economic performance of CCUS networks that connect CO 2 sources and applications or consider carbon policy interventions (52,53).
TEA models for CCUS systems suffer from substantial uncertainties in technology performance and costs, given their low TRL and limited large-scale projects in operation. Because TEA models rely on mass and energy balance data from process simulators, conducting uncertainty and sensitivity analysis, especially for global sensitivity analysis, can be challenging (54). Recent efforts to address this challenge include design under uncertainty by optimizing system variables according to one or several targeted objectives related to the output distribution of cost performance modeled in TEA (e.g., minimizing the mean value of cost or uncertainty range of cost distribution) (54). Another effort to address the challenge is developing surrogate models using ML methods such as Artificial Neural Network to avoid intense computational demands in uncertainty analysis (58). CCU systems have more methodological challenges than CCS. One main issue is the multifunctionality-the question is how to attribute the environmental burdens between CO 2 (which is emitted to the environment in the absence of CC) and primary products of the point sources (e.g., electricity produced by a power plant) (59). The literature shows substantially different GHG results between stand-alone CCU systems using CO 2 from any source and systems integrating CO 2 point sources (60). ISO standards (25) recommend system expansion to avoid allocation. A recent review assessed different methods and recommended substitution, a mathematically equivalent method of system expansion. The substitution method subtracts the environmental impacts of the CO 2 sources without CC, to avoid suboptimal choices of CO 2 sources using other methods such as allocation (59). This method has been used in consequential LCA (CLCA) studies for CCU systems. For example, Thonemann & Pizzol (61) included avoided environmental impacts of the marginal production of conventional chemicals (to be replaced by CO 2 -derived chemicals) and changing marginal suppliers of CO 2 and H 2 in the near and long term scenarios (61). CLCA and attributional LCA are two broad LCA types. Attributional LCA attributes environmental impacts to the functional unit of a studied system, whereas CLCA assesses impact changes in response to a decision or action (62). CLCA is more decision relevant than attributional LCA. However, because the CO 2 market has not been established, determining marginal suppliers that would respond to a small increase in demand could be challenging. Advanced LCA models have been developed to enhance the policy relevance and engineering utility of CCU LCA. For example, Thonemann & Schulte (63) present a prospective LCA model that couples the CLCA of a lab-scale CCU system with a scale-up scheme leveraging flowsheet development and process simulation to understand future impacts at a full operating scale. Hoppe et al. (64) estimated raw material inputs and total material requirement, two common indicators used in MFA, using a cradle-to-gate LCA for CO 2 -based production of methane, methanol, and derived polymers. This insight into resource utilization may be useful for policymakers in understanding the role of CCU in supporting a more resource-efficient society. Another study developed a stochastically determined climate return on investment metric using the carbon footprint results of CCU LCA, which helps identify research and development priorities toward CCU products with the greatest climate benefits even with uncertainty-one of the largest concerns for assessing low-TRL technologies (65).
Recent developments in coupling process analysis, TEA, and LCA have advanced sustainability-informed design and optimization for CCUS. De Luna et al. (66) conducted TEA and a simple LCA for prospective renewable energy-powered electrosynthesis pathways of converting CO 2 and water into commodity chemicals (e.g., diesel and alcohol). This allows simultaneous comparison of the market and GHG-reduction potential of CO 2 -derived and fossilbased chemicals, as well as exploration of how technical performance and market potentials would affect the comparison. Another example ( 67) is an integrated assessment that couples process flowsheet, LCA, and TEA for a CO 2 refinery design that converts CO 2 from direct air capture to methanol and its derivative olefines and aromatics, supporting a holistic evaluation of technical, economic, and environmental performances of different CCU implementation strategies for chemical production in different regions.
Yao et al.
ARjats.cls January 12, 2024 9:49
Biomass and H 2 have been explored extensively as alternative feedstocks for chemical production. These two have overlaps, as H 2 can be made from biomass. This section discusses advanced microand meso-modeling for biomass-based chemicals (including H 2 ) and H 2 from water electrolysis.
Biomass can be converted to value-added chemicals through different platforms (e.g., syngas and C6 sugar) and conversion pathways (e.g., thermochemical and biochemical conversion). Given the large number of biomass feedstocks, platform chemicals, conversion technologies, and chemical products, identifying the optimal pathways is challenging but needed for scaling up. Knowledge-based microscale process analyses have been used to compare alternative pathways and combined with TEA and LCA to support pathway screening (68). One example is an optimization-based process synthesis framework that combines process analysis with LCA and TEA using MINLP (mixed-integer nonlinear programming) to explore the economic and life cycle GHG emissions of converting three biomass feedstocks to liquid transportation fuels through different combinations of biomass conversion units, operational conditions, and upgrading processes (69). In addition, previous studies have used green chemistry and engineering principles to guide biorefinery design and pathway screening (70). ML has been used to overcome both computational and knowledge barriers in sustainable synthesis and design for biomass conversion processes, products, and supply chains. Previous reviews (71,72) found that most ML applications for biomass systems have focused on optimizing either technical performance (e.g., yields of chemical products) or part of the life cycle (e.g., only biomass conversion). A few studies have demonstrated the powerful use of ML to support sustainability assessment for bio-based products. For example, Liao et al. ( 73) combined a kinetic-based process simulation with Artificial Neural Network models trained from experimental data collected from the literature to predict GHG emissions and energy consumption of pyrolysis and steam activation across a wide range of woody biomass feedstock for biochar and activated carbon production (73). Another study trained ML models using data from biomass conversion process simulations and leveraged a knowledge-based decision tree to automate the initial process design of separation and product recovery, allowing for fast evaluation of economic, technical, and environmental performance (74). A recent perspective presents a promising integration of bench-scale digitation (enabled by ML methods such as text mining) and automatic laboratory experiments and multi-scale modeling (from molecular to process level) for LCA and TEA, which can accelerate the identification and optimization of biomass system design for sustainable chemical production (68).
Large-scale biomass utilization impacts land use and ecosystems, which often are not considered in engineering-based models. Land use change (LUC) can be direct (e.g., forest converted to cropland) or indirect [e.g., induced LUC (ILUC) caused by changes in biomass supply and demand]. LUC can result in net carbon emissions or sinks. For instance, a study for bioenergy with CCS identifies LUC as the main factor determining whether bioenergy with CCS is carbon positive or negative (75). Because ILUC is an ongoing research field, different models (e.g., GTAP-BIO and GLOBIOM) may provide different GHG emission estimations (76). In contrast to bioenergy literature, LCAs of biomass-based chemicals rarely have considered LUC. A few studies have considered land requirements based on biomass yields (77,78), but the LUC GHG emissions are not included. Eerhart et al. (79) used GHG emission factors of ILUC associated with corn-based ethanol biofuel for bio-based polyethylene furandicarboxylate and showed significant impacts of ILUC on the GHG mitigation benefits of polyethylene furandicarboxylate compared to petroleum-based counterparts. Waste biomass has been highly recommended, because it does not drive LUC as dedicated crops (e.g., corn) (80) biomass (e.g., agricultural residues) may lead to soil organic carbon losses, which have been explored in biofuel literature (81), but not bio-based chemicals. More research and modeling efforts are needed to understand the large-scale impacts of biomass use for chemical production on ecosystems.
H 2 is used widely in the chemical industry as a feedstock (e.g., ammonia production). Thus, decarbonizing H 2 production and application will benefit the chemical industry. Microscale process analysis for H 2 production typically relies on knowledge-based simulations leveraging the literature and experimental data to assess and optimize process performance for various technology pathways, e.g., electrolysis (82)(83)(84), thermochemical (85), biochemical (86), and electrochemical (84). Data-driven models mainly use ML techniques, and their outputs are process performance indicators (e.g., H 2 yield and energy consumption). ML training data can be collected from lab-scale experiments (87), plant operation (88), simulation models (89), and the literature (90). These data-driven models can be used further for process optimization [e.g., maximizing H 2 yield ( 91)] and uncertainty/sensitivity analysis that is challenging to execute in process simulators (88). Furthermore, they can be integrated with LCA or TEA for fast environmental and economic assessment (92). Compared to traditional models, the benefits of ML models include no need for initial solutions or guesses for the simulation model ( 88), greater ability to model complex phenomena (e.g., convection, conduction, and radiation among the burner-flame guide-steam methane reforming reactors for H 2 production) than equation-based models ( 88), higher prediction accuracy for H 2 production under nonequilibrium conditions (91), and less computational intensity (91,92). Models for H 2 applications have focused on either microscale applications or mesoscale optimization for H 2 networks. For microscale applications, one emerging area is power-to-fuel using CO 2 and H 2 to produce energy products (e.g., methanol, methane, and fuel). These studies integrated the models of CC, H 2 production via electrolysis, and conversion together for a holistic assessment (93)(94)(95). Mesoscale models evaluate the GHG reduction potential or optimize the lowcarbon hydrogen supply chain for industrial uses. For example, Chisalita et al. (96) used LCA to evaluate the environmental impact of ammonia production with different H 2 production scenarios across supply chains (e.g., via steam methane reforming, chemical looping, or electrolysis), and Ibrahim & Al-Mohannadi (97) used MILP to identify the minimum cost of the low-carbon H 2 supply chain networks in industrial clusters and displayed the trade-offs between levelized cost and CO 2 emission reductions. Simultaneously, many MILP-based models for H 2 supply chain network design exist in the transportation sector (98), where the methods can be applied to H 2 used for chemical production.
There are two types of electrification: direct electrification that changes only energy infrastructure and indirect electrification that relies on electrochemistry to synthesize alternative feedstocks (99). Because direct electrification does not change existing chemical processes, previous studies typically began by analyzing existing systems, followed by estimating energy and GHG emission changes after implementing electric facilities. For example, Chen et al. ( 99) developed a knowledge-based process simulation and TEA for a traditional methanol plant and then modified the model for heat pump implementation and estimated energy reduction benefits. Leveraging public databases, Schoeneberger et al. (100) characterized GHG emissions and energy consumption of existing industrial boilers, then estimated the GHG reduction benefits of replacing them with electric boilers. Similar approaches, such as that of Jabarivelisdeh et al. (101), have been used to analyze other incremental process improvements that do not alter chemical processes.
Yao et al.
ARjats.cls January 12, 2024 9:49
Indirect electrification requires substantial changes in chemical process design (93,99). For example, using electrochemistry, methanol can be made from CO 2 through direct CO 2 electrolysis or CO 2 hydrogenation with H 2 derived from water electrolysis, both of which require substantially different design of reactor and feedstock treatment from traditional methanol synthesis via thermochemical conversion of natural gas (93). Thus, it is more challenging to conduct process analysis, TEA, and LCA. The level of modeling details depends on the analysis purpose and scale. For example, process optimization studies often start from reactor or process design and then use LCA and TEA to identify optimal process design (84,102). Other studies have used LCA and TEA in a streamlined fashion without detailed process analysis to identify technical conditions or chemical product targets with the largest potentials in terms of markets and GHG reduction (e.g., 99).
Electrification allows for integration of renewable energy, such as solar and wind, into chemical production (103). On the other hand, the carbon impacts of decarbonizing the chemical industry can propagate and be magnified through chemicals' use as clean energy carriers (e.g., low-carbon H 2 and methanol) (11). Understanding such impacts requires integrated modeling for both the energy and chemical sectors. Macro-level modeling, discussed in the following section, can be helpful.
Various modeling approaches have been used to understand and identify the decarbonization potential of the chemical industry that accounts for its broader impact on supply chains (e.g., energy supply) at a national or global level. This section discusses these approaches with specific application examples and highlights their strengths and limitations.
Industrial ecology models have been used to analyze chemical industry decarbonization at the macro-level. In this section, two categories of industrial ecology models are discussed, namely material flow analyses and bottom-up macro-level models.
MFAs (discussed in Section 2.1) are based on data gathered from domestic sources on flows of materials, import/export information, and statistics or estimates of reuse or loss. MFAs are very useful in illustrating the quantitative life cycle of a particular element; nonetheless, they are stories of the past.
An approach that suggests the future is termed material future scenarios (MFSs). MFSs begin with recent MFAs and build upon those foundations to provide informed stories of one or more possible futures regarding material supply, demand, utilization, and loss over time, often for several decades. A typical MFS addresses a particular geographic scope (country, region, planet); a material group of interest; and a storyline that describes material demand, technology development, societal evolution, and national/international policy over the MFS time period. To allow for various visions of possible futures, MFAs generally involve several scenarios that differ substantially in attributes such as anticipated changes in supply and demand, political stability level in countries that are suppliers or users of the material in question, and perhaps technological or societal evolutions.
Several scenarios specifically related to plastics have been published, with foci on plastics production (104), plastics pollution (105), plastic-related CO 2 emissions (5), and industry transitions toward net-zero CO 2 emissions by mid-century (2, 7). There is general agreement that without significant transition of the industry toward net zero, the CO 2 emissions and impacts on CH15_Art3_Yao ARjats.cls January 12, 2024 9:49 natural ecosystems will at least double by 2050. Studies such as these do not, of course, provide political pathways toward such a transformation, but they do provide the scientific justification for moving to implement those pathways for the good of people, the environment, and the planet. Bottom-up models can be used together with top-down approaches (as hybrid methods). One example is a decarbonization study for the UK chemical sector (109) that used a bottom-up approach to model individual chemical processes and a top-down approach to disaggregate nationlevel steam and electricity consumption data. The decarbonization potential was then analyzed through a bottom-up technology roadmap combined with different technology adoption scenarios. Most previous studies, whether bottom-up, top-down, or hybrid, relied on assumptions of individual technology adoption, which overlooks the possible competition and synergies among different technologies.
Several global decarbonization analyses have considered technology competition using optimization models, such as a technology choice model that addresses the uncertainties associated with technology choices in CLCA (110). A global analysis for plastic life cycles used this method to consider different technology choices (including CCU and biomass) with the objective function of minimizing system-wide GHG emissions (111). Other studies have considered the technoeconomic potentials of different decarbonization choices. For instance, Yang et al. (112) conducted TEA for decarbonizing chemical/petrochemical production based on representative CO 2 refinery processes and their CO 2 emissions, as well as the technical feasibility and costs of typical CC and biomass systems. The authors estimated the best CO 2 emission reduction options under 100 €/CO 2 for the entire sector (112).
Integrated assessment models (IAMs) model the interaction of the chemicals industry with the rest of the economy, particularly the energy sector and its transformation through climate policies. IAMs project future demand for different energy carriers that are used (as feedstock or as energy services) and produced (e.g., hydrogen) by the chemicals industry. Industry decarbonization (113). They also lack regional heterogeneity in technology and process sophistication, which gain importance with the need for different regional solutions in emerging and industrialized economies. Regional or national IAMs (e.g., GCAM-China) may enable a less burdensome way to break out the chemicals sector (114) and model location-specific carbon sequestration potential (115).
Stronger coupling between IAMs and micro-and meso-models in industrial ecology can improve the representation of decarbonization pathways in both types of models. The industrial ecology models can provide IAMs with improved process innovations, recycling, lightweighting, and other material efficiency strategies to broaden the spectrum of emissions mitigation strategies (116). IAMs balance the incorporation of detail in industrial processes against computational and data limitations, as well as structural constraints in balancing macro-level energy statistics and monetary flows. One example of such coupling is the embedding of a plastics supply chain model in the IMAGE IAM (PLAIA), wherein the former provides a detailed representation of plastics production, and the IAM enables a broader set of carbon mitigation options (e.g., circular economy) that can also account for trade-offs with other sectors (e.g., land use in biogenic carbon sequestration) (5).
The IAMs, in turn, provide industrial ecology studies with future projections of energy mixes and prices, decarbonization scenarios in other industry sectors, and socioeconomic futures that drive chemicals demand. Coupling IAM with industrial ecology methods offers opportunities for cross-sectional prospective analysis (116) and supports prospective LCA studies of emerging technologies in chemical production. For example, Lamers et al. (117) presented an open-source framework (LiAISON) to link future energy mix scenarios derived from IMAGE with LCA of emerging technologies that consider future technology improvements and scale-up. The framework was applied to power-to-hydrogen and steam methane reforming, demonstrating how future transitions in power, cement, steel, and fuel industries could reduce the environmental impacts of hydrogen production. Another study explored the country-specific decarbonization potential of biomass-derived drop-in (e.g., fast pyrolysis) and hydrogen-based biofuels with and without CCS in Europe by linking LCA models with the outputs from REMIND IAM, which projects future technology mixes in power, transportation, and industries in a specific policy scenario by 2050 (118) 120), who integrated indirect CO 2 emissions derived from environmentally extended input-output analysis (a top-down IE approach) into the European TIMES IAM model. Both studies found that including indirect GHG emissions affects the results of the optimal power sector mix given by IAM. Because the chemical industry is an important energy demand sector while producing energy carriers (e.g., hydrogen and biofuels), there are many research opportunities in leveraging IAM to better understand the role and impacts of chemical decarbonization technologies in global and regional contexts and cross-sectoral mitigation strategies.
ARjats.cls January 12, 2024 9:49
An emerging research field is absolute environmental sustainability assessment (AESA), which compares the environmental impacts of anthropogenic systems to their assigned nature's carrying capacity based on planetary boundaries (PBs). PBs are defined as the safe operating space for human development and are represented by nine essential earth system processes, including climate change, land-system change, and biochemical flows (121) Most scaling methods used in PB-based AESA rely on sharing principles (e.g., principles based on global and regional environmental impacts, gross value added, and final consumption expenditure), limiting the use of biophysical information that may be more accurate for region-specific assessment (123). Bakshi et al. (124) developed a techno-ecological synergy framework that directly compares human demand for ecosystem services (e.g., GHG emissions or resource uses) with the carrying capacity of ecosystems estimated by biophysical models and data rather than scaling methods. Techno-ecological synergy allows for exploring technological systems co-located with ecosystems and can be integrated with LCA for multi-scale assessment, as demonstrated by previous studies (125,126). Decarbonization technologies affect ecosystems through natural resource utilization (e.g., biomass, lands, and water) and environmental emissions. Spatially explicit ecosystem modeling and data need to be incorporated into LCA and AESA to support region-specific decision making.
This article reviews the advanced and integrated use of engineering, industrial ecology, technoeconomic, and ML models to assess promising decarbonization technologies and strategies in the chemical industry. The selection and combination of these models are purpose and scope dependent. Micro-and mesoscale models support engineering and supply chain design and optimization. Macroscale models provide high-level insights for policy and industrial decision making.
Most process analysis, LCA, and TEA are conducted at the microlevel and rely on the knowledge of process configurations, environmental emissions, and economics, leading to computational challenges and knowledge barriers to scaling up microlevel models to meso-and macroscales. ML addresses this challenge by providing data-driven approaches when knowledge gaps exist or providing computationally cheaper surrogate models. However, previous ML applications were limited to one or two life cycle stage(s) of decarbonization technology (e.g., the CC process only rather than the entire CCUS life cycle) or a single aspect of sustainability (e.g., economic). More research is needed to effectively combine ML with existing systems analysis models to support 3. 16 Yao et al.
ARjats.cls January 12, 2024 9:49 holistic assessment and optimization of decarbonization technologies. MFA has been used for both micro-and macro-level analyses and is particularly useful for circular economy pathways. Two critical issues need to be included in future models for a decarbonized circular economy. One is the carbon in products that dissipate or disintegrate during use; the other are products and packaging lost during or after use. Assessing decarbonization technologies and strategies requires forward-looking, robust models of future chemical production and relevant industrial systems (e.g., energy systems). Previous prospective studies have developed different scaling-up scenarios to explore what-if situations. However, most are limited to chemical production without considering future changes in other industrial systems (e.g., electricity grid), economics (e.g., changes in feedstock supply and prices), and society (e.g., human behavior). LCA and TEA of emerging technologies is a fast-growing field, and recent developments in prospective, anticipatory, and consequential approaches are promising to address this challenge (62,127,128). Application of these new tools requires close collaboration between chemical engineers, industrial ecologists, and economists for tailored modeling strategies to ensure the usefulness of the results with reasonable levels of technical detail and data.
Climate change is not the only grand challenge society faces. Many resources and environmental challenges must be addressed for a more sustainable future. This review includes two recent developments, IAM-and PB-based AESA, that have not been used widely by chemical engineers but provide a means of linking the chemical industry with the environment and society. IAM provides forward-looking, cross-sectional insights on industrial decarbonization beyond the chemical industry, which can be combined with micro-/mesoscale models to understand how technology-specific decarbonization pathways would affect regional and global industries and society. PB-based AESA considers nature's carrying capacity and includes environmental aspects beyond climate change, allowing for a more holistic assessment of decarbonization technologies and strategies in terms of sustainability. More research is needed for seamless integration of these methods to provide useful insights for different audiences. This review does not include social impact assessment; however, methods such as social LCA (129) are under development and should be considered in future assessment of emerging decarbonization technologies.
www.annualreviews.org • Models for Decarbonization in the Chemical Industry